Keel's free crypto portfolio optimizer takes 2 to 10 Hyperliquid coins and works out three mixes from the last 90 days of daily returns: the one with the lowest volatility, the one with the best return per unit of volatility (maximum Sharpe), and an equal split. It plots the efficient frontier, shows each coin and each mix on it, and lists the weights and the correlation between the coins.
What cash earns elsewhere. The maximum Sharpe portfolio maximizes return above it per unit of volatility.
Loading 90 days of daily returns…
Build these coins as an index and backtest the basket with rebalancing and fees.
For each coin, the optimizer starts from its daily closes on Hyperliquid over the last 90 complete UTC days and turns them into daily log returns. From those it takes three numbers: the average daily return, the volatility, and the covariance with every other coin you picked. Only these summaries reach your browser, and the optimization runs there. Averages and covariances are scaled to a year by 365 days.
daily return r = ln(close today / close yesterday)
return a year = mean(r) × 365
volatility a year = stdev(r) × √365Every portfolio here is long only and fully invested: each weight is between 0% and 100%, and the weights add up to 100%. For weights w, expected returns μ and covariance Σ, a mix has
expected return = w · μ
volatility = √(w · Σ · w)
Sharpe ratio = (expected return − risk-free rate) / volatilityMinimum variance is the mix with the lowest volatility. It uses only volatilities and correlations, not returns. Maximum Sharpe is the mix with the highest Sharpe ratio at the risk-free rate you enter. If no coin returned more than that rate over the window, no mix has a positive Sharpe ratio and the optimizer says so instead of picking one. Equal weight splits the money evenly. It uses no estimates at all, which makes it a fair benchmark for the other two.
The optimizer solves these exactly. With at most 10 coins it can try every group of coins the best mix could hold, solve the problem on that group, and keep the best answer whose weights are all positive.
The chart puts volatility on the horizontal axis and expected return on the vertical one. Each coin is a dot. The curve is the efficient frontier: for each return between the minimum variance mix and the best single coin, the lowest volatility any mix reaches. Points on the curve beat everything below them. A portfolio usually sits to the left of the coins it holds, because coins that do not move in step partly cancel each other out. How far left depends on the correlations in the table under the chart.
The frontier is a picture of the past window, not a forecast. Volatility and correlation over three months say something about the next few weeks. Average returns over three months say very little: a single rally can make a coin look like the best asset in the set and put most of the maximum Sharpe weight on it. Correlations also rise in sell-offs, so the spread you see in a calm window can shrink when you need it most. The window is short and the estimates are noisy, so read the weights as a starting point and check how much they move when you add or drop a coin.
The numbers leave out trading costs, funding on perps and the drift between rebalances. To see how a basket held up with those included, build it in the index builder and backtest it. To see how wide the range of outcomes can be for the same strategy, run its trades through the Monte Carlo simulation.
Keel is a strategy builder for Hyperliquid. Describe an idea, backtest it with fees, slippage and funding, and run it live on Hyperliquid’s crypto perps. Start free, and connect a wallet when you’re ready to trade.
It is the set of portfolios that give the highest expected return for each level of risk, measured as volatility. Any mix that sits below the curve has a twin on the curve with the same volatility and more return. The left tip of the curve is the minimum variance portfolio; moving up the curve buys more expected return with more volatility.
Minimum variance ignores returns and finds the mix that moved the least, using only volatilities and correlations. Maximum Sharpe finds the mix with the highest return above the risk-free rate per unit of volatility. Minimum variance depends only on how the coins moved together, so it is usually the steadier of the two from one window to the next; maximum Sharpe leans on average returns, which change a lot.
The last 90 complete UTC days of daily closes, turned into daily log returns. When coins were not all trading for the whole window, it uses the days they share. The result card shows the exact dates. Returns and volatilities are scaled to a year by 365 days, since crypto perps trade every day.
Treat them as a description of the last 90 days, not a forecast. Average returns over three months are mostly noise: one strong rally can put almost all the weight on one coin. The weights change when you add a coin or when the window moves forward a few weeks. Compare them with the minimum variance and equal weight mixes before acting on them.
The portfolios here are long only and fully invested: every weight is between 0% and 100% and they add up to 100%. A coin that is volatile and moves with the others, or one whose return was low, adds risk without enough reward, so the best mix leaves it out. With shorting allowed it would get a negative weight instead.
No. Most coins move with BTC, and in a sharp sell-off correlations tend to rise toward 1, just when diversification is needed most. The correlation table shows the last 90 days only. A mix that looks well spread in a calm window can behave like one position in a crash.