A friend of mine spent most of one bear market averaging into a single losing altcoin, and every time I asked why, I got some version of the same sentence. It has dropped eight days in a row, it cannot drop a ninth. It could, and it did, and the streak eventually outlasted his margin. The reasoning error underneath that sentence has a name, the gambler's fallacy, and I would guess it quietly kills more retail accounts than any of the biases that get more airtime.
The textbook example comes from a roulette table in Monte Carlo where, in one famous session in 1913, black reportedly came up twenty six times in a row. Gamblers lost fortunes that night betting red on the theory that the wheel owed them a correction, and the wheel owed them nothing, because every spin was independent and the odds of another black after twenty five blacks were exactly what they were on the first spin. Everyone reads that story and feels smarter than those gamblers. Then a coin closes red six days straight and the same people start typing due for a bounce into a group chat.
Why price streaks do not self-correct
The intuition behind due is that extreme runs get balanced out, and it feels grounded in real math because in one narrow sense it is. Flip a fair coin for long enough and the proportion of heads does drift toward half. But the law of large numbers works by dilution, an ocean of future flips slowly swamping the early streak, and never by the coin steering back toward fairness. The coin has no memory, so ten heads in a row leaves the next flip at fifty fifty, and nothing anywhere is keeping score on behalf of tails.
Markets are worse than the coin, because price moves are not even independent. An asset that has closed red six sessions in a row is often red for a specific reason. A token unlock is hitting the market, a fund is unwinding, a lending desk is liquidating collateral, the business underneath is deteriorating. Conditional on the streak, continuation can be more likely than a bounce, which flips the gambler's math on its head. Anyone who bought Terra's LUNA on the way down in 2022, reasoning that a coin already down roughly ninety percent had to be near a floor, learned that every new drop is measured from a fresh, lower base and that most of whatever remained could still evaporate. A long red streak describes the past, and it exerts no upward force on tomorrow.
How the fallacy sneaks into position sizing
The expensive version of this bias usually arrives dressed as position sizing rather than as a forecast. Martingale is the casino form, double the bet after every loss so that the first win recovers everything plus one unit. On paper the sequence always ends in profit, and in practice it fails in a casino and in a margin account for the same reason. The doubling grows faster than any finite bankroll, and a streak long enough to break you is guaranteed to arrive if you play long enough. Casinos formalize this with table limits. In trading the table limit is your account equity, and the enforcement mechanism is liquidation.
Averaging down without new information is soft martingale. Each add lowers the average entry, which feels like improving the trade, and each add is justified with streak logic, the idea that so much red has already happened that one green day recovers it all. Run that process for a while and it produces an ugly portfolio property. Your winners stay small because they never give you a reason to add, while your losers compound, so your size ends up largest on exactly the trades where the market has disagreed with you the longest. Accounts built that way bleed slowly and then die suddenly, and the trader remembers every individual add as a reasonable decision at the time.
My rule of thumb here is blunt. If the only new fact since my last buy is a lower price, I do not get to add. A lower price by itself is the market voting against my thesis, and caring more about my average entry than about that vote is how streak logic talks me into betting bigger while knowing less.
When reversal logic is actually valid
None of this means mean reversion is fake. Some of the best documented edges in markets are reversal edges. Short horizon overreaction in equities has been studied academically for decades. In crypto, extreme perp funding rates have historically tended to precede snapbacks, because crowded leveraged positioning supplies the fuel for its own unwind. Liquidation cascades overshoot because forced sellers are price insensitive, the margin engine sells whatever it must at whatever bid exists, and price often retraces once the forced flow stops. Pairs with a structural link, two share classes of one company, a stablecoin against its peg, snap back because arbitrage capital is paid to close the gap.
The difference between those edges and due for a bounce is checkable rather than philosophical. A real mean reversion edge names a mechanism that pushes price back. It defines the trigger, the exit, and the size before the trade exists, and it has been tested across hundreds of instances instead of reconstructed from a few charts someone remembers. Before I take any reversal trade, I run through a short list.
- Can I name the mechanism forcing price back, in one sentence, without using the word due?
- Has the setup been tested over a large sample, ideally hundreds of trades, with results I can actually look at?
- Is my size fixed before entry, with no plan to add just because price moves against me?
- Is there a defined exit if the reversion never comes, either a stop or a time limit?
- Would I open this exact trade right now if I had no position and no memory of the streak?
The last question trips up the most people. A lot of what gets called conviction is an open loss someone is unwilling to realize, dressed up as a theory about balance.
When I feel the pull of a due trade now, I make myself write the rule down as if I were handing it to a stranger. Buy after some number of consecutive down days, exit after a set holding period or at a set loss, fixed size, no adds. Then I backtest it, which is a habit that shaped what we built into Blockcircle, because the honest version of this exercise is humbling. Most streak rules either lose outright or underperform doing nothing once fees and the occasional death spiral are included. A few survive on specific assets in specific regimes, and those are worth trading at a size decided in advance. The rest I let go, which is much cheaper than finding out the ninth red day was waiting for me too.