Prediction methodology

This page explains how every price prediction on TheWeal is computed. It is the longest editorial-policy page on the site for a reason: readers should understand the model before placing weight on the numbers it produces. Same inputs, same outputs, always. Every step below is reproducible from public market data.

What the model is

The TheWeal scenario model produces bear / base / bull price targets for any coin in our coverage universe across eight horizons: 24 hours, 7 days, 30 days, 3 months, 6 months, 1 year, end of the current calendar year, and end of 2030.

The model is deterministic. Same inputs always produce the same outputs. It does not use machine learning. It does not use generative AI. It does not adjust based on author opinion. Every number on a coin page is the algorithmic output of the formulas below, applied to the inputs as of the timestamp shown on that page.

The four components

Each scenario is the sum of four components — momentum, mean-reversion, volatility band, and sentiment overlay. The weights are constant across coins; only the inputs change.

1. Momentum (drift component)

For each coin, we compute a smoothed momentum reading from 7-day and 30-day percentage change:

weekly_drift = 0.7 × pct_change_7d + 0.3 × (pct_change_30d / 4.3)

This is the model’s view of “the rate at which price has been moving recently”. It is intentionally backward-looking. The model carries this drift forward at decreasing weight as the horizon extends — a 24h projection is dominated by recent momentum, a 2030 projection barely uses it.

2. Mean-reversion (anchor component)

For each coin we compute a “reversion target” that price is presumed to gravitate toward over longer horizons. The target is anchored on the coin’s all-time high, scaled down by a quality factor based on market-cap rank:

rank_quality = max(0.15, 1.0 − (rank / 1500))
reversion_target = ATH × (0.7 + 0.3 × rank_quality)

So a top-10 coin has a reversion target close to its ATH; a rank-1500 coin’s target is roughly 70% of ATH. The model blends current price toward this target at increasing weight as the horizon extends.

3. Volatility band (range)

The bear and bull scenarios are not opinion — they are mechanical bands around the base case, scaled by a per-horizon volatility multiplier derived from the coin’s recent realised volatility. Multipliers:

These multipliers were calibrated against historical realised volatility for BTC and a basket of top-50 alts, with the 2030 multiplier deliberately wide to honour the genuine uncertainty over multi-year horizons.

4. Sentiment overlay (nudge)

The base case is nudged by the current Crypto Fear & Greed Index:

• F&G below 30 (fear): base case pulled down by 5–15%
• F&G above 70 (greed): base case pulled up by 5–15%
• F&G 30–70 (neutral range): no overlay

The intuition: in extreme sentiment regimes the market tends to over-react. The overlay encodes that mild contrarian view.

A worked example

Imagine Bitcoin at $73,000 with 7-day change of −6%, 30-day change of −9%, an all-time high of $126,000, market-cap rank 1, and the Fear & Greed Index at 22 (extreme fear).

1. Momentum: weekly_drift = 0.7 × (−6%) + 0.3 × (−9% / 4.3) = −4.83%. Applied to 7-day horizon → $73,000 × (1 − 0.0483) = $69,475.
2. Mean-reversion: rank_quality = max(0.15, 1 − 1/1500) ≈ 0.999. reversion_target = $126,000 × (0.7 + 0.3 × 0.999) = $125,874. Blend weight at 7-day horizon is ~5% — so target nudges base by ~$2,800 upward.
3. Base before sentiment: $69,475 + (some fraction toward $125,874) = roughly $70,800.
4. Sentiment overlay: F&G 22 (extreme fear) — base pulled down 5–10%. Result ≈ $70,800 × 0.95 ≈ $67,300.
5. Bull / bear band: 7-day vol multiplier 0.18 × 90-day realised vol (~62%) × $67,300 ≈ ±$7,500 → bear ≈ $59,800, bull ≈ $74,800.

This is illustrative — the live numbers on each coin page are produced from the exact realised volatility and current sentiment at refresh time, not from approximations.

Refresh cadence

Price-driven inputs (current price, 7d/30d change, volatility) refresh every 60 seconds via CoinGecko. Sentiment refreshes every 30 minutes via Alternative.me. Predictions recompute when inputs change.

Known limitations

The model has documented failure modes. These are not bugs; they are the price of using a deterministic, reactive model.

1. Regime shifts. The model is reactive — it incorporates new information through momentum, but it does not anticipate regime shifts (e.g., a major regulatory event, a black-swan failure). When a regime shift occurs, the model’s base case lags reality for the first 7–14 days.
2. Longer horizons are wide. The 2030 band is intentionally enormous (±350% on the bull/bear multiplier). This is honest: at multi-year horizons in crypto, the genuine uncertainty is enormous and any model claiming otherwise is overconfident.
3. New listings have no history. Coins listed on CoinGecko within the last 30 days have insufficient history to compute realised volatility and momentum. We hide predictions for these coins until 30 days of data are available.
4. Stablecoins are pinned. The model identifies stablecoins by category and pins their forecasts at $1.00 ± 1.5%. We do not generate scenario bands for stablecoins.
5. Memecoins. The volatility and momentum readings on rank-50-and-lower memecoins are extremely noisy. Forecasts on these coins should be read with a fat grain of salt; the bull / bear bands are wide for this reason but readers should not place real weight on the base case for such coins.

Backtesting and audits

We backtest the model quarterly against the prior 90 days of realised price data for the top-100 coins. Backtest results — mean absolute error, hit rate for which scenario (bear/base/bull) contained the actual close — are summarised in a quarterly audit appended to this page (the first audit will publish in mid-2026 once we have a full quarter of post-launch data).

What we do not do

• We do not adjust forecasts to favour partner exchanges, partner projects, or any commercial relationship.
• We do not collapse the bear/base/bull bands into a single point estimate when we report on our own forecasts — readers see the full range.
• We do not republish the model’s output without the methodology link.
• We do not market the model as a trading signal. It is a reference frame for a reader’s own analysis.

If you find an error

If you spot a methodology error, a stale formula on this page, or a coin where the model’s output contradicts the inputs, email methodology@theweal.com. The methodology lead reviews every email within 48 hours and corrects the page on the record if warranted.

Not financial advice. Read the full disclaimer.