Are you taking into account the Canadian tar sands (already included in the EIA C+C data)?

The production is set to increase by 0.5 mbpd in 2007, check this post:

http://www.theoildrum.com/story/2006/10/20/142436/03

Ace has spent huge amounts of time in the existing data, something I have never seen you do.

I've come close to doing some modeling on my own but I have nothing to add of any value over whats been presented. So I'm not sure where or how I could contribute modeling. I've got lots of ideas I'd like to model but it would be a full time job and in general requires data we don't have. The crux of the problem is to do anything beyond whats been done on the modeling front requires data thats not in our possession.

Take my irrational markets and shortages model all you get is almost immediate system collapse so I'm not sure what to model. My other keen area of interest bunker fuel has most of the data behind a paywall and given the data the model looks complex to me. I've looked for free databases of historical price records and even emailed bunkerworld but no reponse.

The model he presented was the most aggressive I've seen so
maybe I was a bit rude in asking but yes I wanted to understand why/how he got it. I'm sorry if it came off rude.
My ant antenna's went up and got burned off by the implications of the graph.

Now with that said we are completely dependent on total liquids and the EROI of these total liquids to keep this house of cards from collapsing early. Outside of NGL's the EROI of the rest is suspect and NGL's probably are offset by the declining quality of the crude. So even if you throw in all liquids your back to Aces curve. BTW this is why I'm not all that interested in all liquids from a EROI perspective with the decreasing quality of crude thrown in they are pretty much neutral or 1:1 on EROI and can be discounted at best say 3:1 and thats probably pushing it.

Given the way I handle none C&C sources Ace's graph caused me to sh%@t my pants which can have the effect of causing a posting to be a bit rude if done before the trip to the bathroom.

Sorry.

Good point. My response to that is "Reserves are not the issue. The issue is how fast you can get it out of the ground. There is an unstated assumption that if oil is "there" that there is no difficulty in getting out of the ground, processed at the well head for water and other impurities, transported halfway round the world, then refined and delivered to you. There are a lot of difficulties, lately with refining, but mostly in getting it out of the ground. We can discuss those difficulties if you want."

At this point the eyes have normally glazed over and they think I am a nutter (probably I am) and they walk away. Most people just do not care. Even educated, aware people, do not care. They do not have the entire matrix of information required to understand the information in their heads. Most people are not aware, they are more concerned with the personal habits of Brad Pitt and Paris Hilton, or football. These are the folks who also whine about "gougers" and being ripped off etc.

Humanity will trip over the edge wondering what dress x or y actress is wearing to the Oscars. Nobody cares.

Or better reserves are known to be 3500 bbls so the discovery curve must match. A bit simpler explanation without a in depth analysis. The prediction discovery is asymmetric. However if you assume that Antarctica can and will be fully exploited you probably do have 3500 barrels. Does anyone know if some people are excluding Antarctica and some are including it. You would have to think their has to be some pretty big oil fields in Antarctica.

Funny the USGS included it with Africa.
http://www.usgs.gov/newsroom/article.asp?ID=636

And they have a lower figure.
I'd have to guess that estimates for the Arctic and Antarctic would be the wild card ?

Perhaps the difference is (at least partly) explained by a difference in what is being measured on the left axis of the two graphs. One is barrels (bbl) and the other is barrels of oil equivalent (boe).

I've pretty much got my economic model done for considering the economic effect of HL/Export Land. I still don't see how to plugin numbers to get estimates.

First and foremost we assume the market is in denial well past when its obvious we have peaked this means the market will be ineffective in pricing oil to include the risk of export land.
It will only react to export land conditions.

This condition of a broken market implies that it will incorrectly allocate resources which of course means shortages will develop so assuming the market is misguided naturally results in shortages from mis allocation of resources. This is a common theme when a resource is over exploited if its renewable or simple exploitation of a non-renewable resource where the expectation is supply is far larger than reality. Humans are by nature bullish so our markets always fall into boom bust cycles its practically intrinsic.

Thats the economic side of the equation. So the primary effect is shortages this leads to the second half and why peak oil is so ugly. What happens when the supply of a resource becomes strained is events that where previously uncorrelated because ample supply acted is a sort of buffer become correlated. In the markets this is known as systematic risk. Even without outright shortages the whole system becomes more and more susceptible to systematic risk as supply and demand drain reserves until outright shortages develop.

The systematic risk or correlation is not obvious but you can see that a shortage in one part of the world prompting panic buying can lead to a shortage somewhere else and another run of panic buying. The how of the correlation or chain of events that cause seemingly unrelated events to actually be correlated is complex.

A simple picture is a spray of hot coals landing on a sheet of paper each burns a hole and if the paper is big enough and the size of the coals small enough and better you can get a fresh sheet the events remain uncorrelated. But in the case of coals landing on paper if two land close together by chance then you can get exponential growth in the burning region they share a common variable in this case proximity.

So for the sheet of paper to go from having small holes burned in it to bursting into flames and completely disintegration we actually have a good idea of the variables that cause correlation its the size of the coals the rate the coals are falling on the sheet and how often two different coals land close enough to share the flame front.

In the case of oil these variables are not as easy to determine but its easy to create any number of correlated positive feedback loops when supply is short to show that shortages and low supplies result in correlation between previously independent events.

Its easy to see from the paper example how a lot of small events can suddenly turn into a big event as the correlation causes coalescence with bigger and nastier feedback loops.

And you can see how hard it is to actually see this coming since its effectively random which the exact chain of events
actually causes the paper to burst into flames.

Back to the model you can however watch for more and more correlation and although its impossible to time you know that past a certain threshold which from other models I set at about a 5% difference between supply and potential demand the chance for a coalescence of correlated events becomes almost 100%. Next of course this model has everything falling apart so fast you really don't have time for the market to respond and try and restabilize the situation.
Forget about trying to find alternatives as punishing shortages cause the economy to grind to a halt.

A major hurricane in the gulf or and attack on Iran will be enough to send our world economy to the bottom. But even if this does not happen the aggregate of correlated small events will eventually result in cascading failure.

The timescale is easy since WT's models shows exports dropping by about 50% within 5 years this economic model of correlated shortages with exponential feedback everywhere has to execute in less than two years once it goes critical.
The minimum is based on simple timescales for the moving of oil and the fact that reserves will not all be at zero in the beginning this gives a lower bound of about 8 months from the beginning of cascading failure until the system cannot recover. I think barring a large event this is to low and its weighted to the high end.

Next of course when does this model start I gave 5% split between potential demand and supply but this is hard to gauge. Going to Exportland agian we can see that two years into Exportland we are well within this 5% region regardless of how demand has adjusted to supply. So two years past when export land begins in earnest is about the most your going to get and then your less then two year before meltdown.

Taking a conservative approach and assuming ExportLand started this year we end right where my predictions keep ending which is we are toast by 2010. So the end of the oil age can be predicted comfortably to be 2010 to 2011 at the latest. So enjoy the next 2 1/2 years.

None of model are valid much past peak. At best 15% of peak production. I doubt they will work much past a 5% decline.
The underlying assumptions are simply not valid past this point.

If you press one of the global warming modelers I suspect they will also admit their models don't hold much past significant melting in the Arctic since its almost impossible to successfully predict through adding a new ocean to the planet.
The models may be accurate but we cannot have faith in them.

All the peak oil models suffer the same problem since oil extraction and usage is a economic/life style choice and concentrated in the middle class of the wealthy countries the actual production after about a 5% depletion has a lot more to do with the economic health of this group than anything else. If they can continue to maintain their consumption patterns and lifestyle then we probably stay close to the current models. This is the weakest business as usual part of these models but inherent through out is that things stay like they are with prices effectively steadily increasing. The chances of this happening esp when you consider the current financial climate are slim to none.

My guess is we will see shortages and major price spikes followed by economic downturns then a bit of recovery rinse and repeat. At each downturn you will shed a certain percentage of the population which will find that it can at best earn survival money. If you take ExportLand and just apply the percentages pretty much directly you have about 10% of the worlds population that used a decent amount of oil becoming minimal participants in the oil economy. Its of course not a flat percentage and it won't be a simple increase. But each year a pretty big hunk of the worlds population is going to have to effectively opt out of the world economy and it won't just be in the poorest of nations. You would expect this to cause above ground factors that won't favor oil production and consumption at anywhere near current levels.

But I hasten to add its impossible to believe any model we have much past peak they will all be invalid fairly quickly I doubt we can correct them quickly enough to actually maintain any predictive value in effect things will change faster than we can get enough data to even make a prediction and correct replace models. I'm not optimistic but I do recognize that we don't really know except that we can be pretty sure the next five years will not be pleasant.

As and aside simply having business loosen their rules for workers and moving to extensive carpooling could easily buy us a lot of time. Also subsidizing moving into the cities. We could even after we wait too late still act in a way that we can convert our society.
At least in America I'd be surprised to see people do such acts since in a lot of cases it would mean actually setting next to a stranger this is intolerable in the US.

Yes.

As soon as i get them up to my photosomething account I will post them.

Our problem is that many exporters, especially the top three net exporters, based on HL, are way, way past the 50% depleted mark, which I think will result in catastrophic net export declines.

The paper that Khebab and I are working on will attempt to quantify the estimated net export decline, at least for the top five. It will not be a pretty picture.

The 60% annual decline rate in UK net exports (a figure I haven't seen any raw data for, but I'm assuming is valid) is presumably because the high rates of production decline (due to the agressive way in which the North Sea fields were exploited) quickly lead to a situation where total oil production was very close to domestic consumption (which has barely increased over the same period).
The 60% was a once-off (and final) drop of course - from now on the UK will be a net importer, barring any substantial new finds.
I'm not sure how this is relevant to the situation of major oil-producing countries with developing economies that are now supplying the world market. Almost all of them are producing far more oil than they consume internally, and it would require at least 20 years at 5% growth in domestic consumption for most of them to end up with zero net exports. Sustaining 5% growth for 20 years is no mean feat, I might add - especially with unstable/authoritarian political regimes. But I would certainly accept that in 20-30 years time there will be extremely little oil left on the world market. The part I don't get is why exports are expected to decline at the sort of rates I've seen you post before (26%?).

My view on the issue is that it is a simple tool that seems to work reasonably well in many large producing basins.

Thats what I said :)

For me, the beauty of the HL technique is that it is intuitively appealing. Indeed, after a few minutes of reflection it is almost intuitively obvious.

I've been messing with parabolas since I made my first reflecting astronomical telescope at the age of sixteen. It is true that for "flat" parabolas a sphere (or circle in two dimensions) is a good approximation, but the "deeper" the curve, the more apparent is the parabolic shape. In other words, you can make a six inch telescope f/12 (seventy two inch focal length) with a spherical mirror which is a good approximation of a parabola. But if you want to make an f/5 (thirty inch focal length) six inch reflecting telescope, the sphere is not at all a good approximation of a parabola.

Because we are looking at "deep" curves (or steep ones) when discussing HL, I do not think that the fact that some parabolas are close to circles or ellipses or hyperbolas is relevant.

Right its only a half model since ants are good citizens and use renewable resources :)

But its close enough since the ants expend their resources extracting their food and digging holes etc etc. On the upside the economics of a ant hill are logistic and what they accomplish given scaling is basically the same thing we do extracting oil. My point is when you scale the problem up and do all the variable assignments between a anthill and humans or really big ants drilling oil wells its a one to one mapping. And thus the logistic model fits. The whole system scales to almost a perfect match with oil extraction. Numerically of course a number of ants are replaced but one human with powerful tools so its not one ant - one human but other than that you can play the game of showing that the two are countably the same problem with a scaling factor. Underlying this is the fact that the energy used is about the same as you scale. Supply/Demand for both product and the means of production maps onto the birth death model for example. So given the ability to do a scaled correspondence the problem is not to prove that oil extraction is logistic in nature but to figure out how too prove its not logistic.

Or at least its logistic to peak and slightly beyond. As you see once you move much past peak its not clear the model holds but it makes sense to move to a collapse model that works for every other bubble situation thats ever been modeled. I'm proposing exactly this as the conditions that take hold post peak so at about 5% decline from the peak you go into the collapse model.

The method of proof is effectively this.

http://polymathematics.typepad.com/polymath/2006/08/countability.html

Or this link.

http://en.wikipedia.org/wiki/Countable

With a bit of induction as Don mentioned.
http://en.wikipedia.org/wiki/Inductive_reasoning

In my opinion the strength of the proof is that you can show
a scaled but very good countability equivalence between a logistic ant hill model and oil extraction. That you have used inductive reasoning to assert are equal in the first place.

Its a pretty dang good proof. And you would be hard pressed to prove its false and even if you found a alternative model
it has to be really close to logistic so your alternatives would almost certainly be variants of the core logistic model.

To be honest at the moment I can't see how these sort of extraction models or resource depletion models can be anything but logistic on the upside given the strong mapping between the dynamics of ant hill with proven logistic behavior. Notice total ant colonies include a discovery phase when they enter a virgin territory so even that in the model so I don't see discovery as outside the logistic model . In effects all factors are in the ant model and its logistic.

If you think about it attempts to disprove peak oil require a sort of magical market trick by Wall Street to save the day or finding Atlantis filled with light sweet oil.

Just curious, why are you so threatened by the Peak Oil concept?

As I have said several times, if anyone follows my ELP advice, and I am wrong, they will have little or no debt, a lower stress way of life, and more money in the bank.

Precisely what are you recommending? That people maximize their spending, commute and consumption?

If the oil drum had existed back then westexas would have been repeating endlessly during 2000-2002 that Saudi Arabia's production was declining just as predicted by HL.

We don't have to imagine. As I noted at the top, Deffeyes observed in the 2001-2002 time frame that the world probably peaked in 2000. However he never backed off his statement that his model showed a peak in the 2004-2008 time frame. He basically observed that the world appeared to have peaked prior to the time period predicted by his model. Note that an observation is not a prediction.

As I noted up top, there are several items that support peak in the 2005 frame--among them high oil prices and the near certain decline/crash of every field that has ever produced one mbpd or more of crude oil.

are simulated examples constrained by geology ?

Hubbert linearization is claimed to be able to predict future production when that production becomes constrained by geology. How is this model supposed to produce a reliable prediction when the data fed into is from a period when production was constrained by a cartel?

I don't think the cartel thing matters. You can have an oil province that is producing with no political contraints but that has some yet-to-be-discovered significant fields. This would be similar to an oil province that had no remaining significant discoveries but was constraining production because of political reasons. An intelligent use of HL should be able to take care of this. All the objections I've seen to HL seem to be based on unrealistic expectations and use of HL as a precise predictive tool, which it is not. This is kind of like the famous 'srawman' argument.

What I like about HL is that it is based on actual historical data and not on the often WAG data of reserves and the even more WAG projections of how oil-extraction projects will develop and come to fruition.

Hi Gil,

Take a look at the production profiles for every UK North Sea field. To save your time, the ones normally described as "giant" would be Forties, Statfjord (which is mostly in Norwegian waters), Brent, Beryl, Piper, Magnus, Nelson, Scott and Cormorant - Thistle turned out to be not-so-giant after all. There is simply no physical, technological, commercial or economic basis for the logistic function.

http://www.og.dti.gov.uk/pprs/full_production.htm

And if someone says that HL works for basins instead of fields, I'll send them right back to the production profile for Azerbaijan - Page 3 of this PDF.

http://www.bp.com/liveassets/bp_internet/globalbp/STAGING/global_assets/...

What I will concede is that exponential decline, or a family of functions of which exponential decline is an end-member, works pretty well for fields where underlying decline in the reservoir (pressure decline, watercut increase or GOR increase) is the main determinant of production rate, i.e. after you've pretty much stopped drilling new wells. There are good reasons for this, founded in physical law (unlike HL). But if you aggregate the production profiles for a bunch of fields with different decline rates, the end result is most definitely not a simple exponential decline curve, even if they're all exponential individually.

Love,

PUD.

I don't think you are stupid; I don't think you are full of crap. I like your comments.

Nevertheless, the HL model is based on pure inductive logic as its basis, with a bit of deduction tacked on to clarify issues.

"Scaling" as I see it is not relevant to the issue, because rather than looking at "scaling" situation, what is going on mathematically is much simpler--to wit, adding together many small HL curves to get one big global one.

So fractals?

it is well known that many obejects in nature are fractally based.

the small ant colony is the same as the large one, if you don't know the scale.

Same with many coastlines. Doesn't matter how far you zoom in, looks the exact same.

Same thing applies here to oil fields, doesnt matter how big the fields are, the cumulative production is logistic.

it's the same with organism growth in cultures, resource exploitation is the same.

Why are you using the term faith as if its derogatory.
I have faith that Quantum Wave Mechanics is true but only the haziest of understanding. I have faith in a lot of physical and statistical concepts. In general its because of a chain of trust either in understanding some of what a person has done so I trust other parts of their work or via faith handed down from a professor or teacher who believes the work and maybe understands it. Chains of faith and trust are a cornerstone of Science. If you have no faith or trust in the competence of some one you will never believe anything they do.

And finally your talking about testing this proof with experimental evidence and a model. Thats a test of the proof the claim that HL works rests of the proof.

If you don't believe the proof YOU test it.
If you don't believe me and have no faith thats a completely different issue nothing I can't do about it.

If you have well completion dates well production logs and equipment men employed directly and indirectly in developing a large field that fits HL over a 20-30 year period send them to us and we can test both HL my proof and a large number of potential models we could also use financial info concerning investment decisions. Given my proof I know the data needed to test it. And I know I don't have it.

Maybe WHT is interested enough that he can think of another data set we do have that would work but this is what I came up with. Maybe well completion dates are enough ?
I tend to think that even if we proved wells completed in a logistic manner it would not be enough for most of the people that doubt HL and I'm sure doubt a physical model might underly the empirical fit. If you think about the data I asked for and read a bit about logistic systems it takes only a few minutes to see that most if not all of the processes involved in developing a field follow logistic equations thus its no surprise the overall system is logistic.

Indeed logistic models are used to model large engineering projects.

Google
logistic regression large engineering projects

I suspect you did not even try to do the proof or even look through the link or google for other very similar systems that all exhibit logistic behavior. I know you did not even try too set up the variables to equate them since you would then know immediately I don't have the data to test the proof and would not have written this.

provide solid and comprehensive evidence, though.

Inductive reasoning is a valid proof so prove me wrong and tell me you did one thing besides writing a quick response.

My only caveat with the shock model is its more complex and thus offers a better fit but its not clear how strong its predictions are in the sense that it has a lot more parameters that need to be set and how does the effect the outcome ?

Since it includes directly discovery it tends to model late large discoveries better than HL which is a problem for HL as its currently written. But with that said double humps tend to
get absorbed and you revert to the mean. So the overall result is similar to doing HL on say the lower 48 + Alaska the central limit concepts prevail in time. So for the current HL when discoveries are truly well in the past i.e 10 years your safe. You can see the problem here is the discovery profile is not included as part of simple HL. If it was then the correct answer is probably to model the fields but the result is the same as doing HL on disjoint data sets. If you broke out the individual fields and summed the URR's instead of doing HL on the composite or overall production data you would then not have a double hump.
So you can see late discovery and doing HL on multiple regions is really the same problem. In short any double peaks caused by discovery exist only because of the way your presenting the data they are not real just a figment of the accounting methods being used I can take any collection of fields and generate double humps. For HL the simple solution is decomposition followed by summing.

Now the big trick with HL that drives everyone bananas and even me is this points used to fit trick. The practitioner of HL is applying some serious filtering and smoothing criteria using the human brain to pick out the parabolic curve. Thats both the power and in some sense the weakness of the method since a least squares fit with smoothing and filtering would eventually give you the same answer but at least we have a mathematical form for the rejection criteria. No problem as long as when you do HL you also show me a plot of the curve and the real data so I can also at least eyeball the fit myself. But we still don't have a rigorous definition of the fitness criteria used.

My answer to this is to at least present the error terms for your HL and for the line say 1% above and 1% below the one chosen so we know you at least have some fitness criteria.

I'm happy to eyeball the curve on top of the data since we are good as seeing fit but its sloppy to not present the error terms and at least some data showing the fitness.

The problem is HL is a trick method like a slide rule from pre computer days. I suspect people objecting to it would also object to how slide rules work. I see how it works but given we have computers why expose the trick instead you can used HL to give the initial conditions for minimization using non-linear least squares methods and get the same answer but hide the trick since it seems hard for some people to understand. I think its cool but I think slide rules are cool :)

WT: I do basically agree with Memmel that your use of contrived mathematical models was a waste of time.

Ok, Robert, let's differentiate.

Your contrived mathematical sets are, like WesTex says, a waste of time and (like I say) a rape of a "perfectly good" model.

ON THE OTHER HAND, like all models, especially because of its shortcut approach, HL has clear drawbacks.

1) Your critique on the applicability to the Texan and KSA situation is VERY WELL FOUNDED. Any partially well-trained statistician looking at the graphs and being aware of the HL method will see the problems in these cases. HL is an eye-baller, like Memmel says.

Therefore, using only HL alone for SA (and the world, for that matter) is irresponsible. It needs to be tested against other models.

Therefore, we need a model against which the HL model can be tested. And, as a further model, the new model would need to be tested against other information.

2) Please show me your complementary model (Have you already done this, do you have a link?). I am somewhat partial to the shock model. I demonstrated my hybrid model, one that does not constantly push URR higher and one that indicates peak a decade or two before it happens.

Texas in 1955:

As you can see, URR is about 66Gb, which is still ca. 2Gb higher than present estimates

Saudi Arabia (all data, please forget all data past ca. 1979 to verify model):

This was the chart I posted in March. If I were to redo the data (a time comsuming aktivity), only using that up to 1979, I come up with a URR of ca. 225Gb. That puts 50% at 112,5Gb which was reached ca. 2006, right?

Why is this model more descriptive than pure HL?
HL uses the last data points to find a conclusion.
Mine asks the question: When was exponential growth no longer POSSIBLE ? (SA doubled production from 2 to 4 and from 4 to 8 very quickly. 16 would have been possible. 32 WOULD NEVER HAVE BEEN IN THE LEAST BIT POSSIBLE)

3) These models are NOT very precise - an error of a couple of years should be accounted for. I don't think even WesTex claimed otherwise. Remember, we're concerned about world peak and not restrictively SA's. The sum of the models insist that we have reached peak/plateau. Do you have a different conclusion?

Thanks!
Dom

Despite claims made that I do nothing I did do a review of HL when Robert posted his first post on the subject.
I sent him my results and he never responded.

1.) HL used correctly is a easy quick way to do curve fitting.
Simply looking at the fit is sufficient to ensure a good fit.
This is true in general with curve fitting looking at the results is a important part of the method.

2.) If you believe the HL anaylsis is in error or want to test other fitting functions a wealth of methods are at your disposal this is and extensive field.
Many are based on the concept of non-linear least squares fitting and their is even a javascript page that allows you ti fit.

http://statpages.org/nonlin.html

This is a iterative minimization method and it gives error terms etc etc.
My response to Roberts assertions is if you don't like HL use this method to fit and its fairly easy to get good error terms.

Next its not clear why we are not doing it this way and treating HL as a first approximation to a minimization routine. But on the same hand your free to do so yourself if you question the fit.

Robert never tried these methods no one questioning HL seems to have the basic mathematical skills in this area to do the obvious steps needed if you question the HL results.

I've laid out a well known approach that produces the information needed to question a curve fitting procedure and it goes unused ?

3.) The fact that people questioning HL don't know these methods and refuse to use them when they are suggested as a minimum first step in criticizing a model and the fact that the second time this has come up non-linear least squares fitting was not suggested even by Robert who has had this information for a while bothers me.

This is not rocket science.

And finally I learned to programing from Numerical Recipes in Fortran and all I did for 4 years was numerical solution for chaotic systems. Your free to decide if I have a understanding of the problem or not. Unless I see the results of some non-linear least squares regressions or other minimization technique I see no reason to even consider the arguments being made that HL is wrong. So at least do the basic steps needed to create the argument against HL.

MY GOD are you people this dense. The data set is assumed to follow the logistic equation and your doing curve fitting.

Why the hell are you talking about debunking HL. Its just frigging curve fitting. I can't believe people keep spreading this FUD.

Curve fitting 101

Any data set can be used to fit to any curve.
Two points make a strait line so any data set with at least to data points can be modeled as a strait line.

Any data set that follows a parabolic like growth peaking and decline curve can be modeled by a large family of functions.
One function that seems to fit well and has some nice mathematical features that may or may not be "real" has been shown to empirically have a good fit for oil extraction.

A infinite number of such curves exist we have explored a few alternatives.

This is just curve fitting its impossible to disprove. Your free to argue that oil extraction does not follow a curve but thats has nothing to do with HL.

Maybe a freshman stats teacher can step in and help me.
This 2+2=5 argument is simply crazy its not even mathematics.

At least please read a introduction to the concept of curve fitting.
http://en.wikipedia.org/wiki/Curve_fitting

This is first year college stuff that you guys are butchering its driving me crazy to see such basic mistakes published and believed.

Astounding I feel sorry for your math teachers.

Lets at least back off and work through the basic principals of curve fitting I'm assuming that we are dealing with people that simply don't have the mathematical background to understand HL and curve fitting. Thats ok lack of knowledge can be corrected.

The concept of disproving curve fitting boggles my mind.

Pitt has many good points. However, I think you've hit on a flaw/frustration with his communication skills. Since he is far from alone in this regard, perhaps we could all remember how communication works.

Communication involves sending, acknowledgment, and validation. When we say something and we have no "valid" response, most of us spend our time repeating ourselves (do you hear me now?). If anyone finds themselves repeating their statements, they should wonder if the other person has any intent to listen, doesn't understand, or could possibly be feeling invalidated.

Pitt, as nice as he can otherwise be, seems to have invalidation issues (puts the bar way too high and isn't very nice regarding this expectation for everyone when it just doesn't matter). Perhaps this makes him feel better about something. It's funny what we unknowingly communicate about ourselves.

Cheer up champ. Don't forget to have fun and make meaningful connections with it.

-Sean.

Memmel,

Consider Great Britain overall as it hit its first peak and began declining. GB is a good example of a "double hump" because of a subsequent large discovery offshore.

Now if I had fitted the HL curve with data up to the first peak, I would have incorrectly predicted a peak in GB production. In other words there are additional factors, in this case subsequent discoveries, that modified the original curve. I believe that the "double hump" problem is partly why Khebab proposed the "loglets" method of analyzing these things.

I agree that the curve itself has little informational value other than how we interpret it but this is a case where an interpretation would have been wrong. This is why I say we need lots of additional sources of data. And we're getting lots of additional sources of data about near peak right now.

I really respect the value of HL but I don't see it as a flawless predictive tool all by itself. It's just not. But it is highly accurate if you interpret the curve in conjunction with multiple other signals. So I guess we basically agree there. The key though is that I can't personally say that just because HL says we're at or near peak that such information, by itself, is trustworthy all alone. For me, HL is just one tool among many that helps me determine that we are at or near peak. When multiple different approaches start to agree, from HL, to HSM, to bottom up analysis, etc., then we have a much higher confidence factor in our subsequent predictions.

Ghawar Is Dying
The greatest shortcoming of the human race is our inability to understand the exponential function. - Dr. Albert Bartlett