I keep coming back to one number when I try to explain why my altcoin book swings so much harder than the amount of money in it should suggest. It is not volatility, though the two are cousins. It is how sensitive an asset is to changes in global liquidity, and once you start measuring it, a lot of otherwise confusing price action stops being confusing. The polite name for it is liquidity beta, and it is one of the more useful lenses I have for thinking about macro without pretending I can forecast a central bank.
The idea is borrowed straight from the old market beta you learned if you ever touched equities. Market beta is the slope you get when you regress a stock's returns against the market's returns. A beta of 1.5 means the stock tends to move about one and a half times as much as the market, in the same direction. Liquidity beta swaps out the benchmark. Instead of regressing against the S&P or against Bitcoin, you regress the asset's returns against changes in a global liquidity measure. The slope you get back is how much that asset amplifies or dampens the tide.
What global liquidity actually means here
Global liquidity is a fuzzy phrase, so it helps to pin down a working proxy before you regress anything. The version most people use is some combination of major central bank balance sheets, bank reserves, and a currency adjustment, usually with the dollar in the mix because a weaker dollar loosens conditions for everyone holding assets priced in it. You do not need to build the perfect index. You need a series that goes up when money is getting easier to find and down when it is getting scarcer, and that is measured consistently over time.
A few practical notes on the proxy, because this is where estimates go wrong before you even run a regression. Use the rate of change, not the level. Liquidity betas are about sensitivity to swings, and a slowly rising level tells you almost nothing week to week. Weekly or biweekly changes tend to be the sweet spot for crypto. And keep the proxy stable across your whole sample. If you splice together three different liquidity definitions over the window, the slope you get is measuring your data-stitching, not the asset.
Why the risk curve sorts betas so cleanly
Here is the part that makes liquidity beta worth carrying around. Assets sort themselves by beta in a way that lines up almost perfectly with how far out the risk curve they sit. Bitcoin tends to have a meaningful liquidity beta but a relatively contained one. Large-cap alts sit higher. Small-cap and long-tail tokens sit higher still, and by the time you get to the truly speculative end, the beta can be several times Bitcoin's.
The mechanism is not mysterious. When liquidity expands, the marginal dollar does not stop at the safest asset. It flows outward, chasing whatever offers more upside, and the assets at the far end have thin order books, so a given amount of inflow moves them more. When liquidity contracts, the same order runs in reverse, and it runs faster, because the marginal seller in a thin market gets no bid. So the amplification is real in both directions, and it is usually a little worse on the way down than on the way up. That asymmetry is the thing that quietly ruins people who size a high-beta book as if the drawdowns will be symmetric with the rallies.
Estimating a token's liquidity beta
The regression itself is boring, which is a compliment. Boring means reproducible. Here is the workflow I actually use.
- Pull returns for the token and for your liquidity proxy over a common window. Historically I want at least a year of weekly data, and more if the token has traded long enough to give it to me. Too short a window and you are fitting noise.
- Convert both to periodic percentage changes. Returns for the token, change in the proxy for liquidity. Align the dates so every row pairs the same period.
- Regress token returns on the liquidity change. The slope coefficient is your liquidity beta. Look at the r-squared too, because a high beta with a near-zero r-squared means the relationship is weak and the number is mostly luck.
- Re-run it on a rolling window, say the trailing few months, and watch how the beta drifts. A token's beta is not a constant. It climbs in speculative phases and compresses when the token starts trading on its own narrative.
Two failure modes worth naming. The first is a token that is too young or too illiquid to have a trustworthy series, where the regression hands you a confident-looking slope built on a handful of real observations. Treat those betas as rumors, not measurements. The second is lead-lag. Liquidity does not hit every asset on the same day, and if the token reacts with a delay, a same-period regression understates the true sensitivity. Testing the proxy lagged by a period or two often tightens the fit.
Using beta for sizing and rotation
Once you have a beta you trust, the first use is defensive. If you size every position by dollar amount, a basket of high-beta alts carries far more liquidity risk than the dollar figure implies, because they all move together when the tide turns. A rough correction is to scale position size inversely with beta so that each position contributes something closer to equal risk. A token with twice the beta gets roughly half the allocation for the same risk budget. It is crude, it ignores correlation between your holdings, but it beats equal-dollar sizing by a wide margin, and it stops the far end of your book from quietly becoming the whole trade.
The second use is rotation timing, and this is where it gets interesting. If you have a read that liquidity is turning easier, the high-beta names are where the amplification pays you, so you rotate out the curve. If your read is that liquidity is tightening or just stalling, you rotate back toward lower-beta assets and cash, because those are the positions that bleed least when the reverse order runs. You are not predicting the turn with any precision. You are deciding how much amplification you want to be holding when it comes, which is a much more honest question.
The honest caveat is that beta is backward-looking and regime-dependent. It tells you how an asset behaved relative to liquidity over your window, not how it will behave next quarter, and the relationships break during genuine stress when everything correlates to one. So I treat the number as a dial for how much exposure I want, not a signal to trade on its own. When I am watching this across a basket, having the returns, the liquidity proxy, and the rolling betas sitting in one place matters more than any single estimate, which is a chunk of why we built the scorecard and backtesting side of Blockcircle the way we did.
The habit that has served me best is simply to know the beta of everything I hold before I need it, not after. Estimating it on a quiet week costs an hour. Estimating it during a liquidity shock, when the far end of your book is down twice what you expected, costs a lot more than that.