What Is Averaging down?
Averaging down means buying more of an asset after its price has fallen below your entry, which lowers the average entry price of the position. For example, buying one unit at 100 and another at 80 gives an average of 90, so the position breaks even at 90 instead of 100. For a short position the mirror image is adding as the price rises.
Automated versions are sold as “DCA bots” on crypto trading platforms: after an initial order, the bot places further “safety orders” each time the price drops by a set step, then closes everything at a take profit above the average entry. This is not the same as dollar cost averaging in the investing sense, which buys a fixed amount on a time schedule whatever the price does. A DCA bot adds only to a losing position, and when each safety order is larger than the last it is a martingale strategy.
Averaging down improves the break-even price at the cost of a larger position in an asset that is moving against you. It pays when prices revert, which is why it appears inside mean reversion strategies, and it loses most when a fall continues: the position is largest at the bottom. Without a limit on the number of additions and a stop loss or maximum loss for the whole position, the tail risk is that of short volatility.
Discretionary traders are also drawn to it for psychological reasons. Investors tend to hold losing positions too long and sell winners too early, a pattern called the disposition effect (Shefrin and Statman 1985; Odean 1998), and adding to a loser postpones realising the loss. Averaging down as a planned rule with a fixed budget, as in scaling into a value position, differs from adding to avoid admitting a mistake.
Vaults using this approach: martingale-like and mean reversion vaults.
Literature and references:
Hersh Shefrin and Meir Statman, The Disposition to Sell Winners Too Early and Ride Losers Too Long: Theory and Evidence, Journal of Finance 40(3), 1985.
Terrance Odean, Are Investors Reluctant to Realize Their Losses?, Journal of Finance 53(5), 1998.
See also