Every stablecoin depeg I have watched closely looked obvious in hindsight and confusing in real time. The obvious part is that the price left one dollar. The confusing part is that by the time the price actually moves, the interesting information is already an hour old. If you are watching the peg itself, you are watching the last thing to break, not the first. The stuff that breaks first lives on-chain, in the pools and the redemption queues and the issuer's own wallets, and most of it is visible to anyone willing to look before the crowd does.
I want to walk through the signals I actually watch, in roughly the order they tend to fire, and the rough thresholds that make me start reducing exposure. None of these are magic numbers. They are the levels where the odds stop being in my favor, which is a different thing from a prediction.
Pool composition skew is the earliest tell
Most large stablecoins have a deep liquidity pool somewhere, usually a Curve-style pool paired against other stables. In a healthy pool the balance sits close to even, each asset holding roughly its target share. When holders start quietly getting nervous, they do not announce it. They swap the shaky stable for the safe one inside that pool, and the pool tilts. The suspect asset climbs toward a larger and larger share of the total while the safe legs drain out.
That skew is the single earliest number I trust. A pool that normally sits near balanced and drifts to something like two thirds of one asset is telling you that informed money is rotating out while the peg still reads a clean dollar. The reason this works is mechanical. These pools price with very low slippage near balance, so the first wave of sellers can exit at basically par and the screen price barely flinches. The imbalance moves before the price does because the pool is designed to absorb the first sellers without repricing. By the time skew gets extreme, the next sellers eat real slippage, and that is when the visible discount finally appears.
My rough rule: if a major pool crosses somewhere around 65 to 70 percent weighting toward the stable I hold, and it keeps climbing instead of mean-reverting within a few hours, I treat that as a real signal rather than noise. Balanced pools breathe. Sustained one-directional drift is the thing.
Secondary-market discount and where it shows up
The next signal is a genuine discount on secondary markets, and the location matters more than the size. A stablecoin can trade a fraction of a cent under a dollar on a thin venue all day and it means nothing. What I care about is a persistent discount on the deepest venues, the ones where you could actually move size. When the biggest pool and the biggest centralized order books both show the same discount at the same time, that is not a fat finger. That is the market repricing redemption risk.
A useful mental split:
- A discount under roughly 20 to 30 basis points on deep venues is usually just friction, arbitrage lag, or someone dumping into thin liquidity. I note it and move on.
- A discount that holds past half a percent on deep venues, for more than a brief window, means arbitrageurs are choosing not to close the gap. That is the important part. Arbitrage exists to erase these discounts, so when it stops working, the arbitrageurs have decided the risk of holding the asset through redemption is worth more than the free spread.
- A discount that keeps widening rather than stabilizing is the one that turns into a run. Stable discounts can sit for a while. Accelerating ones rarely reverse cleanly.
The trap here is treating all discounts as equal. A one-cent gap on a sketchy pair is background hum. The same gap where the arbitrage should be trivial and nobody is taking it is a warning.
Redemption flow and issuer reserve movements
For fiat-backed coins, redemptions are where intent becomes visible. You cannot see the bank wires, but you can often see the on-chain half. Large burns of the token clustering in a short window, mint activity going quiet while burns accelerate, tokens flowing into the issuer's redemption addresses in size. A burst of redemptions is not automatically bad. Issuers process redemptions constantly and that is the system working. What I watch for is redemptions running hot while new mints dry up, because that asymmetry means demand has flipped to one side and nobody wants to be the marginal new holder.
Reserve movements are the harder, slower signal, and they matter most for collateral-backed and crypto-backed designs. For an on-chain collateralized stable you can usually read the collateral directly. Watch for the collateral ratio drifting down toward its danger zone, the backing assets themselves losing value or liquidity, or large collateral withdrawals from the protocol. A stablecoin backed by something volatile inherits that volatility with a delay. When the collateral is falling and the peg is still holding, the peg is the lagging indicator, not the safe one.
The failure mode that catches people is reflexivity. If a coin is partly backed by its own sister token or by assets that fall precisely when the coin is under stress, the collateral evaporates exactly when you need it most. That feedback loop is why some depegs go from a small discount to near zero in a single session. If you cannot answer the question of what backs this and what happens to that backing under stress, you are holding a black box and you should size it like one.
Turning the signals into a routine
Here is the actual stack I run, cheapest checks first:
- Watch the composition of the one or two deepest pools for the stable. Sustained skew past roughly two thirds is the first flag.
- Cross-check the discount on the deepest pool against the deepest order book. Both showing the same gap past half a percent, and holding, is the second flag.
- For fiat-backed coins, watch burn-versus-mint flow and redemption-address inflows. Redemptions hot while mints stall is the third flag.
- For collateralized coins, track the collateral ratio and the health of the backing assets. Ratio sliding toward its liquidation zone is the third flag for that class.
One flag is a reason to pay attention. Two firing together is my line to start trimming, and I would rather be early and slightly wrong than on time and stuck in the redemption queue. The whole point is that exiting near par is nearly free when the pool still absorbs sellers, and brutally expensive once the discount is on the screen for everyone. You are trying to be in the first wave of sellers, the wave that gets out at a dollar, not the third.
You can wire most of this up yourself with a block explorer and a pool dashboard, and honestly that is a fine way to start, since building it by hand teaches you what normal looks like for each coin. When I want the pool skew, the cross-venue discount, and the wallet flows sitting in one view instead of six tabs, I lean on Blockcircle for the on-chain side, but the thresholds matter more than the tool. Pick your levels before you are stressed, write them down, and let the pool tell you what the price has not admitted yet.