A coin round-trips over a month and ends where it started, and the 3x long token tracking it is down double digits. When this happens to someone for the first time they usually assume fees, or a shady issuer, or a bug in the product. In most cases none of those apply. The token has delivered three times each daily return, exactly as documented, and the word daily carries the whole problem.
A 3x leveraged token gives you three times the return of the underlying from one rebalance to the next, which in most products means one day. It does not give you three times the return over your holding period. Those sound like the same promise, and over a single day they are. Over any longer stretch they diverge, and in choppy markets they diverge badly, always in the same direction, against the holder.
What the daily rebalance actually does
To keep leverage pinned near 3x, the product has to adjust its exposure every day, or whenever leverage drifts past a threshold, depending on the issuer. After an up day the position has grown relative to its collateral, so effective leverage has dropped below 3x and the fund buys more exposure. After a down day leverage has risen above 3x, so the fund sells some off. Put those together and the fund is mechanically buying after prices rise and selling after prices fall, every single day. In a market that alternates up and down days, it is buying local tops and selling local bottoms on your behalf, forever, with perfect discipline.
In a smooth trend the same mechanism works for you. Buying more after every up day in a market that keeps going up is just compounding into a winner, which is why these products can genuinely beat three times the total move during a strong clean run. The rebalance has no opinion about anything. It compounds whatever path the market takes, and most paths in crypto involve a lot of chop.
A worked example in a flat market
Take a coin at 100. Day one it rises 10 percent to 110. Day two it falls about 9.1 percent, back to 100. A spot holder is flat. Now run the 3x token through the same two days. Day one it gains 30 percent, so 100 becomes 130. Day two it loses three times 9.1 percent, roughly 27.3 percent, and 130 becomes about 94.5. The underlying went nowhere and the token lost around five and a half percent in two days.
No fees in that example, no funding, no bad fills, just arithmetic. Percentage gains and losses are asymmetric, a loss needs a bigger gain to recover from it, and leverage magnifies the asymmetry while the daily reset locks it in. Run that same round trip ten times, about a month of sideways chop, and the token has lost roughly 40 percent while spot is unchanged. This is volatility drag, and it is the whole reason long-term holders of these products get ground down without any single day looking catastrophic.
Estimating the bleed before you buy
You can put a rough number on the decay in advance. For a token with leverage L, the expected daily drag in a sideways market is approximately L squared minus L, divided by two, times the square of the daily volatility. For a 3x token that works out to three times the daily variance. It is an approximation, but it is close enough to be useful, and the worked example above matches it almost exactly.
So suppose the coin has been moving about 4 percent a day, which is not unusual for a mid-cap in a lively stretch. Square that to get 0.16 percent, multiply by three, and you are bleeding roughly half a percent a day if the market goes nowhere. Compounded over a month that is somewhere around 13 or 14 percent, before the management fee most issuers charge on top. At 2 percent daily vol the drag drops to about 0.12 percent a day, which still adds up to a few percent a month. Because the drag scales with the square of volatility, a doubling of vol quadruples the decay, and crypto vol regimes can double fast. The same token that behaved tolerably in a quiet month can fall apart in a loud one.
My pre-purchase check, which takes about five minutes:
- Pull the last 30 daily closes for the underlying and compute the standard deviation of daily returns. That is your realized daily vol.
- Square it, then multiply by L squared minus L over two. That is your estimated daily decay in chop.
- Multiply by the number of days you honestly expect to hold. If that figure rivals the move you are hoping to capture, stop.
- Read the issuer's rebalance rules. Some rebalance on a fixed daily schedule, some only when leverage drifts outside a band. The band versions decay less in mild chop, but the core problem never goes away.
When the token works, and when a perp is the better tool
There is a real use case, and it is narrow. If you expect a sharp directional move over the next day or two, and you do not want to manage margin, a leveraged token is a reasonable blunt instrument. It cannot be margin-called the way a perp position can, since the deleveraging is built into the rebalance, and in a fast market that is worth something. The caveat is that a move of roughly a third against a 3x product inside a single rebalance window would still effectively zero it, which is why issuers run emergency intraday rebalances during crashes, and those tend to happen at the worst prices of the day.
For anything longer than a few days, or any market grinding sideways, a perp is usually the better tool. A perp holds whatever notional you opened until you close it, so there is no daily reset and no path dependence from rebalancing. Your carrying cost is funding, which is visible on every exchange, typically a few basis points per interval on majors, and occasionally paid to you rather than by you. In exchange you take on liquidation risk and the obligation to actually watch your margin. My personal rule of thumb is simple. Expected hold over three days, or realized daily vol over 3 percent, take the perp. If both are under those lines, and you have genuine conviction the move will be fast and clean, the token is acceptable.
When I want to check myself, I backtest the equivalent perp position over a similar window in Blockcircle and set it next to the leveraged token's actual historical path. The gap between the two curves is the decay, made visible, and it is usually bigger than I guessed. Most of the time the comparison talks me out of the token, and I have come to treat that as the comparison doing its job.