How We Estimate an Asset's Future Price

There's no perfect way to predict where an asset's price is headed — every method has blind spots and none of them will nail the future. But a rough estimate is still useful: it gives you a feel for what your investment horizon might look like.

We use three panels, each coming at the problem from a different angle:

• Logarithmic regression

• The EV dice roll

• Geometric expected gains

Logarithmic Regression

Pick an asset — say Bitcoin. On its price chart, the price is bounded by three lines plus the live price:

• Predicted fair value — the model's average

• Predicted high

• Predicted low

• Current price

Why logarithmic? Assets grow explosively when they're small and slow down as they get big. A $1M asset only needs another $1M to double. A $100M asset needs another $100M — same 2x, but 100x the money. Because that gets harder as the asset matures, a log curve fits the shape better than a straight line.

And since the model is really just a line in log space, we can extend it forward and read off the fair value years down the road.

Fair values (live, calculated from today)

Fair value isn't the current price — it's the model's guess at a "normal" price. The real price floats above or below it. Trading below fair value means you can expect stronger returns; above it, weaker ones. The percentages show how much the fair value itself grows over time.

Highest prediction

Lowest prediction

The predictions for the logarithmic regression model also includes a highest prediction and lowest prediction, predicting the range the asset can be in. In these charts for example the 2 year prediction of bitcoin can be somewhere between the lowest value 66,552 and a highest value of 229,037 with a fair value or average value of 123,462

Caveat: the fit suits Bitcoin and ETH well — they hug the corridors. Gold and the S&P wander in a much wider band and rarely touch the edges, so trust their lines less.

Reference Charts:

https://www.lofiledger.com/dashboard?tab=price

https://www.lofiledger.com/dashboard?asset=Eth&tab=price

https://www.lofiledger.com/dashboard?asset=Gold

https://www.lofiledger.com/dashboard?asset=S%26P500

EV Dice Rolls

This builds on the regression model's risk levels. Every price maps to a risk score based on where it sits between the low band, fair value, and high band.

The idea: look at Bitcoin's entire history, find every day it sat at a given risk level, then check what happened next — one day later, one year later, two years later — and average those returns. So if Bitcoin is at 0.15 risk today, we ask: historically, when Bitcoin was at 0.15 risk, what did the next two years look like?

Based on this examples Bitcoin, for example, is expected to have a percent gain of [UPGRADE REQUIRED] % in the next two years, which would put it at [UPGRADE REQUIRED] from the current price at the price of 63,490.10

The key difference from regression: the dice roll starts from today's actual price and applies a historical return, so the current price matters. Regression ignores the current price entirely and just extrapolates fair value.

Caveat: this assumes history repeats — that past behavior at each risk level still holds. And since the risk levels come from the log model, it inherits that assumption too.

Reference Charts:

https://www.lofiledger.com/dashboard?panel=expectedValuesDiceRoll&tf=2y

https://www.lofiledger.com/dashboard?panel=expectedValuesDiceRoll&tf=2y&asset=Eth

https://www.lofiledger.com/dashboard?panel=expectedValuesDiceRoll&tf=2y&asset=Gold

https://www.lofiledger.com/dashboard?panel=expectedValuesDiceRoll&tf=2y&asset=S%26P500

Geometric Expected Gains

The geometric mean is the average compounding rate — the single growth rate that would've carried an asset from its starting price to today.

Quick example: if something gained 5%, 6%, and 10% over three years, the average isn't the simple mean. It's the cube root of 1.05 × 1.06 × 1.10 = 1.0697 — about 6.97% a year.

We compute this per asset two ways: across its full history, and on a rolling one-year window to catch recent behavior. Then we average both, since the rate drifts as the asset rises and falls. Project that rate forward and you get a future price.

Price predictions: current price compounded at each annualized geometric mean.

Caveat: this assumes one fixed compounding rate, forever. Real assets slow down as they grow, so the further out you project, the less it holds — which is why these numbers drift away from the other two methods over long horizons. Rolling values are also biased toward recent movements, but the latest method is slightly biased in crypto as the gains early on tend to be more explosive. Overall for crypto, the latest or the average of the rolling values can give us an accurate idea on what to expect, while the latest is useful for the S&P and Gold.

Reference Charts:

https://www.lofiledger.com/dashboard?panel=expectedGains&tf=monthly
https://www.lofiledger.com/dashboard?panel=expectedGains&tf=monthly&asset=Eth

https://www.lofiledger.com/dashboard?panel=expectedGains&tf=monthly&asset=Gold

https://www.lofiledger.com/dashboard?panel=expectedGains&tf=monthly&asset=S%26P500

Summary:

Logarithmic Regression is more useful for Bitcoin and Eth, while the Geometric expected gains are more useful for S&P and gold, however these are unreliable for long term predictions as they do require assets to move exponentially, for longer timeframes this may be slightly unreliable.