Learn · open method
The MPT Score: the Whole Method
Method mpt/1.0 · published
Modern Portfolio Theory, from Harry Markowitz's 1952 paper, is about risk and return together: what a portfolio earned for the risk it took, and how much of that risk its holdings cancel out by not moving in lockstep. The MPT score uses the parts of the theory that can be measured reliably from history (volatility, correlation, drawdowns and realised returns against a public benchmark) and is careful with the part that can't (expected returns). It is free on every plan, and every step is published here.
The six factors and their weights
| Factor | Weight | What it measures | How it becomes points |
|---|---|---|---|
| Diversification benefit | 20% | The diversification ratio: the weighted average of your holdings' volatility divided by your portfolio's true volatility (which counts how they move together). | 1.0 (no benefit) = 0; 1.6 or more = 100. |
| Risk efficiency | 20% | The volatility of the least-volatile mix of the same holdings, divided by yours. | 1.0 (you are at that mix) = 100; 0.5 or less = 0. |
| Return per unit of risk | 20% | Your Sharpe ratio and Sortino ratio, each compared with a 60/40 blend of SPY and AGG over the same days, using the 3-month Treasury bill rate. | Sharpe gap +0.25 = 100, -0.50 = 0; Sortino gap +0.35 = 100, -0.70 = 0; the two averaged. |
| Drawdown | 15% | Your worst peak-to-trough fall divided by the 60/40 blend's over the same days. | 1.0 or less = 100; 2.5 or more = 0. |
| Best mix of your holdings | 15% | Your Sharpe ratio divided by the best Sharpe the same holdings could reach. Expected returns are anchored to CAPM (Treasury bill plus beta times a 5% equity premium), 75% of the way, and to each holding's own history for the other 25%; no holding may exceed 25% (or 1/N) of that mix. | 1.0 = 100; 0 or below = 0. The mix itself is never shown: it is an estimate, not a recommendation. |
| Correlation concentration | 10% | The value-weighted share of pairs of holdings whose correlation is above 0.8. | 0% = 100; 50% or more = 0. |
The data behind it
Daily adjusted closes on the days every holding and both benchmarks traded, over up to three years. The score needs at least 252 shared trading days for holdings that make up 80% of your value; below that it says so instead of guessing. Covariance uses the Ledoit-Wolf (2004) shrinkage estimator, which steadies the estimate when history is short. The rate is the 3-month Treasury bill on the scoring date.
Why expected returns are handled so carefully
On a few years of history, the “best” mix of a set of holdings tends to pile into whatever happened to rise most, which rewards concentration after the fact. So the best-mix factor anchors expected returns to CAPM rather than to raw history, caps any single holding at 25% of that reference mix, and counts for only 15% of the total. The mix itself is never shown.
What it is not
It is not a forecast and not advice. Past volatility and correlation change. It is one lens; the index-investor score is another, and Moneyta's Moneyta Score a third.
Sources: Markowitz (1952), “Portfolio Selection”; Sharpe (1966), “Mutual Fund Performance”; Sortino and Price (1994); Ledoit and Wolf (2004), “A well-conditioned estimator for large-dimensional covariance matrices”; Choueifaty and Coignard (2008) on the diversification ratio.