--- title: "alpha is the residual, not the river" canonical_url: https://ensurance.app/guide/alpha-is-the-residual-not-the-river markdown_url: https://ensurance.app/guide/alpha-is-the-residual-not-the-river.md subtitle: change the factors and the alpha moves. the cooling water does not category: nature-finance --- # alpha is the residual, not the river *change the factors and the alpha moves. the cooling water does not* Investment alpha is the return your model could not explain. You regress a return stream against the market, or against the market plus the factors you chose, and whatever the fit does not account for gets a Greek letter. In a risk meeting the word is precise: alpha is a residual, and a residual is defined by the model that produced it. That precision is the useful half, and it is worth conceding in full before the turn. A residual tells you something real. It says that after you have paid for market exposure, the stream did something the market did not. Whether that something is skill, luck, or a driver you forgot to include is the entire question, and it is a fair question to ask about a token sleeve. Now hold a second object next to it. Palo Verde Generating Station, near Tonopah, Arizona, does not sit on a river. It cools on treated municipal effluent piped west from the 91st Avenue Wastewater Treatment Plant in Phoenix, under a purchase agreement for up to 80,000 acre-feet a year that runs through December 31, 2050 (APS 8-K, April 23, 2010). That effluent is what the cities drank first: a blend of Salt and Verde river water, Colorado River water through the Central Arizona Project, and groundwater. No factor model for a digital-asset sleeve has a column for that water. So whatever the model leaves over, the water is not in it. :::johnson **alpha is what is left after the market. the river was never in the market.** [for investors →](https://ensurance.app/solutions/investors?from=guide&topic=correlation-not-the-river) ::: Correlation, alpha, and a hedge are statements about return streams. The cooling water a desert plant buys under contract, the rivers the cities drink, and the aquifer under that valley exist whether or not those streams move together. Ensurance is how that living system gets funded. It is not the statistic. --- ## what alpha is a residual of ### what is investment alpha? **Investment alpha** is the portion of a return stream not explained by the market or by the factors a model includes. Fit a return series against a benchmark. The slope is beta: how much the stream moves with that benchmark. The intercept is alpha: the average leftover once the slope has done its work. The period-by-period leftover is the residual. In a risk meeting, alpha and residual name the same idea — what the model did not explain. Change the benchmark and the intercept moves. Add a factor and it moves again. Alpha is a statement about a model as much as a statement about a manager. The number on the slide is only as good as the list of things the model was allowed to see. ### how is alpha different from beta? **Beta** is the part of a return the model explains: the stream's sensitivity to the market, or to whichever factors you chose. It is exposure you can buy cheaply, because anyone can hold the market. **Alpha** is the part the model does not explain, and you pay for it because, by construction, you cannot get it from the index. | piece | what it is | who gets paid for it | |---|---|---| | **beta** | the stream's sensitivity to the market or the factors you chose | almost nobody; that exposure is available in an index fund | | **alpha** | the intercept: the average leftover after beta | the manager, if it persists and is not a beta you have not named | | **not in the model** | drivers the factor list never included | nobody in the model; they show up as alpha, noise, or a loss nobody attributed | --- ## the factor you left out ### can a model create alpha by leaving a factor out? Yes. If a real driver of returns is missing from the model, the model cannot attribute that driver's contribution to beta, so it lands in the residual. The residual gets called alpha. Nothing was earned; something was omitted. The history of factor investing is largely this story told in reverse. Return that a one-factor model called alpha was later explained by size, then by value, then by momentum, then by profitability and investment. Each new factor moved a slice out of the skill column and into the exposure column. What looked like alpha was beta to something the earlier model had not named. You might be thinking: fine, that is an argument for a longer factor list. It is, for drivers that already trade. A longer list of price factors still only sees prices. Water futures and weather derivatives exist, and each is a price for a contract in one market. No factor model for this token sleeve has a column for the condition of the Salt River system. A traded water index is not that condition. ### what does a token model actually see? A factor model for a digital-asset sleeve is built out of return streams: broad crypto beta, equity beta, a rates sensitivity, a liquidity or funding term, maybe momentum. Each of those is a time series of prices. Fit the sleeve against them and you get a beta to each and an intercept. That intercept is what a desk calls the sleeve's alpha. Nowhere in that regression is the condition of the Salt River system: snowpack, forest cover, sediment after a fire, or the cities' supply blend. A hot summer in central Arizona can move operating names that drink from that system. The token regression has no column for the condition. It can have a column for a water price. Those are different objects. So the river cannot show up as alpha. Not because it is efficiently priced, and not because anyone has shown it to be uncorrelated with anything, but because it was never a candidate factor in the first place. --- ## the river is a cause, not a residual On one side, a sleeve whose residual moves every time you change the factors. On the other, a plant whose cooling water arrives or does not, regardless of what any model attributes to what. Follow the water back. Palo Verde buys effluent from the Valley cities. The cities produce effluent because they deliver drinking water, and they deliver drinking water out of the Salt and Verde system, the Central Arizona Project, and the aquifer. Upstream of the whole arrangement, Salt River Project and the cities already spend on parts of that system: reservoirs, forest health, watershed work. That spending sits in a general fund and a rate base. It is not in anybody's alpha. You might be thinking: an unpriced risk is exactly where alpha comes from, so why not trade the names that depend on the water? Someone might. A trader who correctly anticipates a shortfall in the effluent chain and positions in the utilities, industrials, or municipal credits along it could earn a residual that a coarse model calls alpha. That is a trade on the condition of the river. It changes the trader's return. It does not change the river. The second objection is the honest one: if the dependency can be modeled, model it. Build a water-stress factor for the operating book and see what it explains. What you will find, if the factor works, is a shared physical cause that several names load on at once — a beta, not an alpha, and one the token sleeve almost certainly does not carry either way. Naming the factor does not fund the thing the factor measures. It only tells you which positions drink from the same pipe. That is the boundary. Alpha is a statement about what a model left over. The living system is what the model was never built to see. One is a number that moves when the factors move. The other is a watershed that responds to snowpack, permits, and stewardship, and to nothing in the regression. --- ## what alpha is not ### what is alpha not? Alpha is not a return. It is the unexplained part of a return, and it only exists relative to the model that failed to explain it. Alpha is not proof of skill. A persistent residual can be skill. It can also be a factor nobody has named yet, a liquidity premium, a tail risk that has not paid out, or luck over a short window. Alpha is not stable. Change the benchmark, the factor list, or the window, and the intercept moves. The same track record produces different alphas under different models, and each is correct about its own model. Alpha is not a supplier. No residual delivers water to a plant gate, power to a node, or soil to a field. It measures an outcome of price behavior; it does not cause anything physical to happen. And a certificate is not alpha. A **specific ensurance certificate** is the record of a hold on one named natural asset. Proceeds go to that place's account and fund work on the system. It gives the holder no water right, no share of anyone's acre-feet, and no rights against an issuer. It is not structured or offered as a security, and nothing here is an offer to sell one. The Salt River system upstream of the effluent contract is a place to look at — [salt-river.basin](https://ensurance.app/salt-river.basin?from=guide) — not an intercept. A certificate is not a return stream, and nobody should put it in the alpha column. Our stage, plainly: specific certificates are live, volumes are small, and the investor and capital-provider pages are where a desk starts. We do not run a correlation model, an alpha product, or a hedge fund, and nothing here is investment advice. --- ## what a desk can do with the distinction Keep the residual. It is a real question, and a token sleeve should be asked it like any other position: after the factors you chose, what is left, and is any of it a beta you have not named? Then ask the second question, which no factor list will answer for you. Which operating names in the book share a physical input at a place — a permitted water source, a river a city drinks, an aquifer under a fab or a data center — and what happens to them together in a dry year? The answer is not an alpha, and it is not supposed to be. It is a dependency. If the answer points at a named watershed, a longer factor model will not fund the system. What a hold on that system looks like is the last post in this series. This post only draws the line: the residual is not the river. The pillar post in this series covers what the correlation number itself can and cannot tell you: [what asset correlation actually measures](/guide/what-asset-correlation-actually-measures?from=guide). [see how investors use this →](https://ensurance.app/solutions/investors?from=guide&topic=correlation-not-the-river) --- ## the series 1. [what asset correlation actually measures](/guide/what-asset-correlation-actually-measures?from=guide) 2. [alpha is the residual, not the river](/guide/alpha-is-the-residual-not-the-river?from=guide) 3. [a hedge has to name the risk](/guide/a-hedge-has-to-name-the-risk?from=guide) 4. [an alt bucket is not a driver](/guide/an-alt-bucket-is-not-a-driver?from=guide) 5. [the correlation outside the matrix](/guide/the-correlation-outside-the-matrix?from=guide) --- ## sources [APS Form 8-K, April 23, 2010](https://www.sec.gov/Archives/edgar/data/7286/000095012310038881/c99652e8vk.htm) — Municipal Effluent Purchase and Sale Agreement: up to 80,000 acre-feet a year of treated effluent from the 91st Avenue Wastewater Treatment Plant to Palo Verde, term through December 31, 2050.