Speculation on this topic generally focuses on some supposed "peak" in production, a point where petroleum is getting harder to extract at a rate faster than our extraction technology is improving. The peak is the point where petroleum production reaches its maximum, after which production will forever be less than that peak. Knowing when that peak is to occur - or has it already occurred? - should help us plan ahead, or so one common brand of thinking goes.
It seems obvious from a mathematical perspective that petroleum production must have a peak at some point. If one accepts that the time interval over which production occurs is bounded both in the past and in the future, at the very least by the emergence of the planet earth from the debris left by the last local supernova and its disappearance in the thermonuclear fire of the next local supernova, and if we represent production as a series of numbers, perhaps the daily tallies, then some number from that finite set must be the largest. Perhaps that maximum occurs multiple times, but if the tallies are given with sufficient precision the likelihood of a repeated number diminishes to insignificance.
What we're interested in, though, is not the mere existence of that peak, but its timing. That, I want to argue here, is chimeric. At a very practical level, the time of the peak may very easily jump across decades, depending on accounting details buried deep in the footnotes. Once you start to play along, the game is so easy that you will surely be able to generate many more accounting tricks than I will suggest here and you will be able to cause the peak to jump across decades all by yourself. But let me make the first few moves, in case they're not immediately obvious.
First, the tally intervals need to be established. This involves setting both their duration and their exact start points. For example, a yearly tally might start on January 1 but perhaps some other date is preferable. Given that production is spread across the globe and undertaken by diverse organizations, the exact start and stop points of the intervals might vary by time zone or by fiscal years in use. The final tally for e.g. 2008 might be the sum of production by all the producing organization, each in their individual fiscal 2008, many of which might be mostly in calendar year 2007.
In any year, the production rates day by day will surely not be constant. One year could have a total production less than some later year, but the first year might easily have a day whose production total exceeds that of any day in the later year.
The sequence of production tallies using one rule for intervals might have several similarly large numbers but only one true maximum. By changing the accounting rules, the numbers will all change a bit, easily by several percent. There won't be an exact correspondence between the two sequences of numbers, because they are defined over different intervals. But if there are two numbers very close in size but quite far apart in timing, a small adjustment can bring the smaller number up a few percent and the larger number down a few percent, so the maximum production tally can be shifted to a very different timing in the sequence.
Petroleum product has ramped up to current levels over almost 200 years. It seems likely enough that production rates are not likely to double yet again from current levels. Nor does it seem likely that production will plummet in just a few years. We are almost certainly faced with decades of fluctuating production, a bumpy plateau, before a slow decline sets in. There isn't a clear line between the plateau and the decline - such a distinction can only be made precise by arbitrary choices of accounting rules and curve fitting parameters. The exact intervals over which production is tallied, that is a first such accounting rule by which such neat distinctions as peaks and plateaus can be manipulated.
Another crucial facet of production accounting is gross flow versus net flow. What is most interesting to society at large is the net output of useful petroleum products such as gasoline, kerozene, etc. that the petroleum industry makes available for end uses such as transportation, heating, etc. Gross production would be the total volume of material extracted from oil fields etc. There can be a lot of variation in the amount of end product made available from whatever fixed amount of gross production. Some crude petroleum may be lost in transportation - more will be lost, generally, from remote fields. Crude petroleum is a class of material which covers a lot of variation - not every barrel will yield the same amount of refined product.
Some really difficult accounting comes into play when considering internal use by the petroleum industry of refined product. However much gasoline etc. is used by the petroleum industry in exploring, extracting, transporting, and refining operations, that much gasoline was not provided for end use by society at large. But how can one really draw a line where the petroleum industry starts and stops? What about the refined product consumed by e.g. the steel industry in producing the tools and structures it provides to the petroleum industry? Should we deduct that petroleum to determine net petroleum production?
As petroleum gets more and more difficult to extract, the petroleum industry will consume more and more resources per barrel of oil. However the exact accounting is performed, a steady gross production will be yielding a declining net production. But the details of the accounting will generate significantly different sequences of net production tallies, which will likely give peaks at very different times.
The exact timing of peak petroleum production is an artifact of arbitrary details of statistical analysis and accounting. The exact timing of the peak is useless for any kind of planning purpose. What is much more useful is a rough forecast of production, something like average production rate decade by decade for the next century. Today we are producing around 85 million barrels per day of crude. In the decade 2030-2040, is this likely to increase to 90 or 95 million or more, will it hold roughly steady, or might we see a decrease to 75 million barrels per day or even less? These are meaningful questions, however difficult to answer. Guessing the timing of peak production is a waste of time.