The word Reserves may mean many things and is widely misunderstood. It may mean the estimated production from the beginning to the end of the field life (ultimate reserves) or that of what is left to produce. The estimates too may refer to what is “proved” so far, hence with a high degree of certainty (called also reasonable) or to what can be “expected” from the field over its full production life (called “mean” or “expected” value). World oil production is also badly defined, equalling 65 Mb/d of “conventional” crude oil or 75 Mb/d when including all liquids. Obviously, the goal should be to forecast the future production of all liquids since the demand is that of these liquids. The starting point is to estimate conventional oil, later adding NGL, unconventional oil and refinery gains, which together comprise supply.  
In most countries other than the USA and FSU, Conventional oil depletion can be studied on the basis of estimated “Mean” Reserves backdated to the year of discovery. We assume that the declared 2P values (Proved + Probable) are close to the “Mean” values because some 2P values will grow but some will decrease, as others that correspond to hypothetical fields, will disappear of the discovery list. The US, following the Securities and Exchange Commission (SEC) rules, report only Proved Reserves, which in plain language means Proved So Far, accordingly being subject to strong reserve growth. In the FSU, The A+B+C1 reserves of their classification are generally thought to correspond with Proved + Probable Reserves, but actually also includes Possible Reserves. Both US & FSU reported reserves need to be corrected to a “mean” value to be useful in forecasting.
Modelling future production is straightforward in countries not constrained by political policy, a situation that does not exist so far with OPEC and the FSU. The Creaming Curve, plotting cumulative mean discoveries versus the cumulative number of new field wildcats, is a powerful tool. Another method is to plot the percentage of annual to cumulative production versus cumulative production, and extrapolate the trend. Future production can also be modelled with one or more bell-shaped curves. The bell-shaped curve was invented by K. Hubbert in the 50s and successfully applied to the production of the 48 US lower states, which are bell-shaped curves similar to a normal curve. In other regions with several E&P episodes, one has to use several such curves, each one related to its discovery pattern.
Oil production in the world outside OPEC and the FSU can be readily modelled with such a bell-shaped curve, despite the recent surge of deepwater production and suggests a peak of about 32 Mb/d during this decade. Modelling the FSU has to recognise the over-production of the late 1980s and the under-production during the 1990s following the collapse of the Soviet Union. What has been under-produced in the 90s or under-developed in the Soviet era is now creating a new cycle. The main contributing regions will be East Siberia and Sakhalin, as well as new discoveries in the Caspian, together giving a peak of about 10 Mb/d in 2010. Modelling OPEC production is more questionable, but it could peak in 2020 at between 40 and 45 Mb/d. 
Combining the three crude oil models for OPEC, the FSU and the rest of the world gives a world ultimate recovery of 2200 Gb. with a peak at about 80 Mb/d between 2010 and 2020. Non-conventional oil and other liquids, with an ultimate of about 800 Gb, could peak in 2050 at around 20 Mb/d. The combined model for all liquids gives a peak of 90 Mb/d around 2015.
Annual world oil production as a percentage of remaining “mean” reserves has increased from 1% in 1960 to 2.5 % in 2000. If demand was to rise to 115 Mb/d by 2020 it would require the finding of some 300 Gb of additional reserves, for which there is little evidence. Hence our conclusion that demand is likely to be constrained by a combination of high prices, recession, and political action.