I keep coming back to the same nagging thing about plain dollar cost averaging. It buys the exact same amount when an asset is 5 percent off its high and when it is 60 percent off its high, and something about that feels lazy. The whole pitch of DCA is that you get more units when things are cheap, but a fixed contribution treats a shallow pullback and a full-blown liquidation cascade as the same event. They are not the same event. So the obvious idea is to make the contribution respond to how far price has fallen, or to how violent the tape has gotten, and buy more precisely when everyone else is puking.
That obvious idea is right, roughly, and also a great way to blow up your cash reserve if you are not careful. I have run enough backtests on this to know the intuition is sound and the execution is where people get hurt. Let me lay out how I think about it.
The two flavors that actually work
There are really only two levers worth pulling here, and they are correlated but not identical.
The first is drawdown-scaled buying. You define a baseline contribution, say your normal weekly or monthly amount, and you multiply it based on how far the asset sits below a reference high. That reference can be an all-time high, a rolling 90-day high, or a moving average you trust. The deeper the drawdown, the larger the multiplier. In backtests this tends to beat fixed DCA in assets that mean-revert around a rising trend, which is most of crypto and a good chunk of equity indices over long horizons. It struggles, badly, in things that just grind down forever and never recover, because you keep feeding a corpse.
The second is volatility-inverse sizing. Here you scale the buy relative to recent realized volatility, buying larger when vol is low and smaller when vol is spiking. This sounds backwards at first because the big drops happen when vol is high, and you would think you want to buy those. But volatility-targeting is really about smoothing your risk contribution over time so no single chaotic week dominates your cost basis. The honest read from backtests is that vol-inverse sizing improves your risk-adjusted return and your drawdown profile more than it improves your raw return. It is a comfort trade as much as a performance trade, and comfort is underrated because it is what keeps you executing the plan.
You can combine them, and I usually do, but I weight drawdown heavier for accumulation and use vol only as a dampener so I am not going all-in during a week where the asset is swinging 15 percent a day.
A tiered multiplier table you can automate
Here is the part you actually wanted. This is a drawdown-tiered multiplier applied to your baseline contribution. Pick a reference high, measure the current percentage below it, and look up the multiplier.
- 0 to 10 percent below reference: 1.0x baseline. This is just normal DCA. Nothing is on sale yet.
- 10 to 20 percent below: 1.25x. A normal correction. Lean in a little.
- 20 to 35 percent below: 1.75x. This is where real discounts start historically.
- 35 to 50 percent below: 2.5x. Uncomfortable territory, which is the point.
- 50 percent or more below: 3.5x, capped. Beyond a cap you are just guessing, and you need reserve left for the next tier down.
Notice the multiplier grows faster than linearly in the middle and then flattens with a hard cap. That shape matters. If you let the multiplier run away at extreme drawdowns you will exhaust your cash exactly when the asset is most likely to keep falling, which is the classic knife-catch. The cap is not timidity, it is what keeps you solvent enough to buy the tier below if it comes.
For the vol dampener, I take that multiplier and shave it when realized volatility is running well above its own recent average. A simple version: if 30-day realized vol is more than roughly 1.5x its 90-day average, multiply the whole thing by 0.7. It slows you down during the genuinely chaotic stretches without turning the strategy off.
The cash buffer is the whole ballgame
This is the part nobody wants to hear. Dynamic DCA is only meaningfully different from fixed DCA because it holds back capital to deploy in drawdowns. If you are already 100 percent invested with every dollar you have, there is no dry powder to scale up with, and the multiplier table is just decoration. The buffer is the strategy.
So you have to decide, in advance, how much reserve you are carrying and over how many tiers you can fund the escalation. Run the arithmetic before you ever start. Ask yourself: if this asset falls into the deepest tier and stays there for a year while I keep contributing at 3.5x, do I run out of cash? If the answer is yes, your multipliers are too aggressive or your reserve is too thin. I would rather cap the top tier at 2.5x and never miss a buy than run 4x and go dry after two months.
The buffer also has an opportunity cost you should acknowledge honestly. Cash sitting in reserve is cash not compounding, and in a market that just rips upward without ever giving you a deep drawdown, plain fixed DCA that stays fully deployed will beat you. That is the trade. You are paying a small drag in the melt-up scenario to get a much better cost basis in the ugly scenario. Whether that is worth it depends on the asset and your read of it, and anyone who tells you dynamic DCA strictly dominates has not backtested a bull market that never corrected.
Guardrails so it does not become knife-catching
The failure mode is always the same. Someone reads the multiplier table, sees the asset down 60 percent, dumps their entire reserve at 3.5x, and then watches it go to down 80 and has nothing left. Rule-based buying feels safe because it is systematic, but a bad rule executed with discipline just loses money faster. A few guardrails I would not skip:
- Use a rolling reference high, not the all-time high, at least as a sanity check. If an asset is down 70 percent from a top it printed once during a mania three years ago, that top may never be relevant again. A 200-day or 90-day high keeps you honest about what "cheap" means now.
- Add a trend filter as a kill switch. If price is below a long moving average and that average is itself falling, cap your multiplier at 1x or pause escalation entirely. You are trying to distinguish a drawdown inside an uptrend from a structural decline, and the two demand opposite behavior.
- Cap total deployment per tier, not just per buy. Spread your escalation across several contributions inside a tier rather than firing the whole tier's budget on the first day price touches it. Drops overshoot. Let them.
- Write down your reserve exhaustion point and check it every contribution. The moment you are within one tier of running dry, you stop escalating and revert to baseline, no exceptions.
If you automate this, and you should because the whole point is to remove your emotions from the deepest tiers where you least want to buy, keep the logic boring. A lookup table, a volatility multiplier, a trend kill switch, and a hard reserve floor. That is the entire system. The temptation is to make it clever with signals and overrides, and every override I have ever added in a live account made it worse, because the overrides always fire in exactly the moments the systematic rule was built to survive. Set the table, fund the buffer, respect the cap, and let it run.