Forecasts are the easy half
Most attention in trading goes to the prediction — which coin, which direction, when. But the account statement is decided by a second, quieter skill: how much, and what you do when it goes wrong. Two traders can take the identical signal and one ends the year up while the other is wiped out, purely through sizing and exits. This guide is the practical framework — no astrology, just arithmetic.
Drawdown math you cannot unlearn
- Losing 10% requires +11% to recover. Losing 50% requires +100%.
- Three 20% losses in a row — entirely normal in crypto — leave you at 51% of your starting equity.
- A single 80% loss — one full-size altcoin position through a bear cycle — requires +400%, which historically takes years.
The asymmetry means the game's first rule is survival of the tail scenarios, not maximization of the base case. Every rule below exists to respect that asymmetry.
Position sizing: two workable rules
Fixed-fraction risk — decide the maximum fraction of equity you're willing to lose on one idea (1–2% is the professional default; 3% is aggressive). Divide that by the distance from entry to invalidation. A trade risking 5% of price to its stop with a 1% account risk budget gets a position of 20% of equity; the same setup with an 8% stop distance gets 12.5%. Notice sizing is derived from the structure of the trade, not from conviction.
Volatility scaling — target a constant exposure to market movement instead of a constant dollar amount. If a coin's daily swings are 3× Bitcoin's, hold ~⅓ the notional. This is why forecast confidence bands matter for sizing: the band is the volatility estimate. Wide band → smaller position, mechanically.
Both rules beat conviction sizing for one reason: they are self-correcting. After a losing streak, fixed-fraction positions automatically shrink, slowing the bleed exactly when your judgment is most likely compromised.
The Kelly lesson: use the brake
Optimal-bet mathematics (the Kelly criterion) says edge over odds determines ideal size — and its actual lesson for traders is how much smaller than optimal you should bet. Full Kelly produces violent drawdowns on realistic estimation error; the standard professional compromise is half-Kelly or quarter-Kelly. In crypto, where edge estimates are uncertain and fat tails are the norm, "quarter-Kelly" is not timidity, it is correct Bayesian humility about your own numbers.
Correlation: the portfolio illusion
Spread capital across 12 altcoins and you may hold one trade. In stress regimes, alt correlations to BTC run 0.8+ and to each other nearly as high — diversification inside one asset class, long-only, is mostly cosmetic. Practical corrections:
- Count factor exposure, not ticker count. 12 alts long = "crypto beta, long, levered" — size it as one position.
- Diversify across regimes instead: some capital in spot-only core, some in strategies with different failure modes, some dry powder in stables.
- When every position in your portfolio is showing the same P&L color, that is not skill or luck — that is one factor. Respect it, because it works in reverse without warning.
Stops: structure, not feelings
Stop placement has one honest principle: put it where the trade thesis is wrong, not where the pain is tolerable. If the thesis is a breakout holding a level, the stop belongs under the level, full stop. If your forecast's 85% band extends to −8%, a stop at −3% will be hit by noise the model already priced. And a stop's job is not preventing losses — losses are the business cost — it is preventing one loss from being the last one.
A workable portfolio shape (core-satellite)
- Core (60–80%) — BTC/ETH-weighted, low-touch, rebalanced on a calendar, not on emotion. This is your exposure to the asset class's long drift.
- Satellites (20–40%) — the ideas with a live edge: AI-scanner picks, forecast convergence setups, each sized at fixed-fraction risk with structural stops.
- Reserve — stablecoin dry powder sized to let you average or redeploy after drawdowns rather than sit through them as a spectator.
The satellites are where tools like FARAZ's forecasts and scanner earn their keep; the core is what keeps you solvent while they do it.
The pre-trade checklist
- What fraction of equity does this risk if the invalidation hits? (Know before entering.)
- Where is the thesis wrong, structurally? Is that level consistent with the forecast band?
- Is this position a new factor or a repeat of an existing one?
- What is my max portfolio drawdown line where I de-risk everything and stop trading for a week?
Answer those four in writing and you are ahead of most participants in this market. For how forecast outputs feed the sizing math, see how AI predictions work; for the record-keeping that keeps you honest with yourself, see hash-verified track records.