Subject: Predictability
A long time ago, the dark ages around 2006, I created a spreadsheet to estimate the future trajectory of Berkshire's share value and price. It's a stochastic/ensemble model, so it can calculate the range of likely outcomes as well as the average among them, or estimate the chances of any particular outcome. It assumed that true trend value would rise at a gradually falling rate, and that price would be sort of a bell curve distribution above and below that value trend, and that there would be multi-year stretches of over- or under-valuation.

I stumbled across the version that has had no updates since book value and price at end 2008, 17 years ago. (FWIW, the estimate of on-trend rate of nominal value growth per share for 2026 was 9.18%)

Unfortunately this was before I switched everything to predict returns above (and therefore separate from) inflation, so it's in nominal price terms, and it wasn't really saying what the real gain might be. Still.

Anyway, its average forecast for market price at end 2025 was about $614,600. The actual number was $754,800. Too high by 22.8%, or nominal 1.2%/year faster than expected. Still, given the length of time, I'd consider that prediction pretty respectable. The model estimated that a price of at least $754,800 had a 14% chance of occurring. The actual market price when the forecast was made was $94,750 per share.

Any excuses I can make for the error? Well, P/B was higher than the 20 year average at end 2025. If it had been at the average, the error would have been only 0.6%/year. Also, a part of the model that wasn't really used (some columns related to SWR) had inflation of 2%/year pencilled in. Since we had an inflation spike, at a stretch one could make the excuse that the estimate was too low partly because of that : )

Not a perfect forecast by any means, but close enough that it added some value for planning purposes. I kind of feel sorry for people who invest in those things that provide no idea of how much money they'll make!

Jim