PUBLIC EVALUATION · PRICE DISTANCE · PROTOCOL V1

How close the forecast came.

A price projection is not well described by a binary label. Here we measure percentage distance from the observed price, both at the expected time and inside an explicit timing margin.

916,227prices evaluated
0.213%median error at target time
0.020%best median error within ±2 h
0.014%best median error within ±3 h

By horizon

Percentage distance, not a pass mark.

Error is the absolute distance between forecast and observed price, divided by the forecast price. Lower is better. The last column compares the model with leaving the initial price unchanged.

HorizonSampleTarget timeBest within ±2 hBest within ±3 hWithin 0.25% in ±3 hVersus unchanged price
1min 78,057 0.110% 0.005% 0.004% 96.4% 0.035 pp worse
10min 111,271 0.196% 0.017% 0.014% 90.3% 0.024 pp worse
20min 62,081 0.256% 0.015% 0.013% 90.6% 0.022 pp worse
25min 51,189 0.278% 0.020% 0.013% 90.5% 0.022 pp worse
35min 36,389 0.323% 0.033% 0.016% 85.7% 0.020 pp worse
40min 32,092 0.348% 0.042% 0.020% 81.7% 0.019 pp worse
45min 29,550 0.378% 0.049% 0.026% 78.4% 0.018 pp worse
50min 24,305 0.363% 0.052% 0.027% 78.2% 0.018 pp worse
55min 20,085 0.342% 0.052% 0.028% 78.0% 0.017 pp worse
75min 14,557 0.388% 0.072% 0.040% 72.9% 0.017 pp worse
65min 16,705 0.365% 0.061% 0.033% 75.9% 0.017 pp worse
70min 15,700 0.368% 0.062% 0.034% 74.7% 0.017 pp worse
80min 13,524 0.417% 0.084% 0.046% 71.3% 0.015 pp worse
85min 12,572 0.444% 0.093% 0.051% 70.1% 0.016 pp worse
90min 12,613 0.438% 0.103% 0.057% 68.8% 0.016 pp worse
5min 186,403 0.114% 0.013% 0.010% 93.9% 0.019 pp worse
15min 84,263 0.212% 0.014% 0.012% 91.5% 0.020 pp worse
30min 48,485 0.305% 0.026% 0.012% 90.9% 0.019 pp worse
1h 21,171 0.355% 0.057% 0.034% 75.2% 0.016 pp worse
2h 10,670 0.474% 0.137% 0.085% 64.2% 0.012 pp worse
3h 12,049 0.733% 0.246% 0.202% 52.8% 0.014 pp worse
4h 8,373 0.964% 0.401% 0.296% 47.7% 0.010 pp worse
6h 3,103 0.872% 0.470% 0.374% 44.0% 0.010 pp worse
1d 2,411 2.113% 1.375% 1.290% 24.1% 0.003 pp worse

The current cohort begins on 10 July 2026. Percentages update as more horizons mature. Earlier rule sets are not blended in.

Method

What each number means.

01

Forecast price

It is calculated from the price seen by the model and its forecast percentage change. Both are stored before the horizon ends.

02

Target time

We compare with the latest valid close at the target time using the finest available market series. Its timeframe and observation gap are published as well.

03

Two-hour and three-hour windows

We find the observed close nearest to the target before and after the expected time. A window never reaches back before the forecast was created.

04

Unchanged-price reference

We repeat the same calculation as if price had stayed at its initial anchor. This shows whether the forecast added anything beyond natural proximity.

How to read it

Proximity and timing are different dimensions.

A wider window makes proximity easier

That is why two and three hours are separate and the error at the exact target time remains visible.

Touching the price does not prove timing accuracy

The JSON also includes distance to candle highs and lows and the time offset of the nearest close.

Horizons need context

A 0.5% error may be small for one asset and large for another. Symbol, volatility, sample and horizon should be read together.

This is not trading performance

It does not include spread, slippage, fees or an entry and exit rule. It evaluates the price projection, not account profit.

Check the price, the timing and the reference.

The aggregate data is public and can be filtered by symbol or horizon.