Data
Where the prices BlissPoint uses come from, what their limits are, over which period estimates are made and which rate enters the Sharpe ratio.
Where the prices come from
The historical prices BlissPoint uses come from Yahoo Finance and are retrieved with the open-source library yfinance. They are closing prices adjusted for dividends and splits: without that adjustment a dividend payment would look like a loss and would distort the calculations. BlissPoint does not own this data and has no commercial relationship or licensing agreement with Yahoo Finance.
What this implies
Availability, accuracy, completeness and timeliness of the data depend entirely on the source. A security may be unavailable, have a shorter history than expected, or contain missing values; in those cases BlissPoint reports it and, if necessary, leaves it out of the calculation. The service can also become temporarily unreachable: when that happens the analysis does not run and an error is shown, with no partial results.
What we do with it
The prices are used only for statistical processing: daily returns, volatility, correlations and the weights of the two portfolios. The series stay in memory for the duration of the calculation and in a short-lived cache that avoids requesting the same data repeatedly; they are not stored in the database and are not made downloadable. BlissPoint does not redistribute market prices and does not offer them as a product: what you see on screen are the percentages computed for your portfolio, not the raw data.
How many requests we make to the source
The minimum necessary. Each analysis downloads the series for the requested securities once; the result stays in an in-memory cache for fifteen minutes, so two people analysing the same securities within the same quarter of an hour count as a single request. The example page is computed once every six hours, not on every visit. If a download fails it is retried at most twice, with an increasing wait, and then it stops. Each registered user can run a limited number of analyses per hour. There is no background process downloading data, and no historical archive that grows over time.
Which period is used, and recently listed securities
The analysis uses the period in which ALL the chosen securities have a price: that is the common window, and expected return, volatility and correlations are estimated on it. The dates need not match one by one — Milan closes on 26 December and New York does not — so a day enters the calendar when the exchanges the basket is listed on were open. Which days an exchange was open is stated by that exchange's OFFICIAL calendar, not by the securities you picked: it is the only way the answer does not change when the basket changes. Adding a security on a venue that is already there does not move the others' figures by a decimal, not even if that security missed some sessions. The few missing prices are carried forward from the last known quote within a few days. A recently listed security does shorten the window for everyone else, and that cost is declared rather than hidden: the «Data and method» panel shows completeness, that is how much of the available history actually went into the estimates, and when it is low there is a young security inside. If the common window falls below the minimum needed to estimate anything, the security that shortens it most is set aside and the page says which one and why: nothing disappears silently. For one day there was a version that estimated on a ragged panel, with every security bringing its whole history; it was correct in theory but was withdrawn together with three other changes that, taken together, had made the frontier and the portfolios unrecognisable. A simple method whose limits are visible beats a sophisticated one that moves the results for no visible reason. Annual figures come from multiplying the daily mean by 252 and the standard deviation by the square root of 252: it is the convention used throughout Markowitz theory, and it is always the same whatever the basket, so two analyses can be compared. VaR is the only figure that needs days on which the securities really were all there together, because it is a percentile of observed portfolio returns.
Which risk-free rate goes into the Sharpe ratio
The Sharpe ratio is the return in excess of the risk-free rate divided by volatility, so that rate is no detail: it is half the numerator. BlissPoint uses the official overnight rate of the currency — €STR published by the European Central Bank for the euro, SOFR published by the New York Federal Reserve for the dollar — fetched when needed and kept in memory for a few hours. Both rates are published on an ACT/360 basis and are converted to a calendar year by multiplying them by 365/360 before they enter the Sharpe ratio: without that step the rate would be understated by 1.4%. The value used, the source and the reference day are written next to each portfolio's Sharpe ratio: anyone comparing two numbers must be able to see whether they were computed with the same yardstick. When the basket mixes currencies, the returns are already measured in a composite unit — the weighted sum of price changes expressed in different currencies — and the risk-free counterpart of that unit is the average of the rates of those same currencies. The right weights would be the portfolio's, but those are known only after optimising while optimising needs the rate: the shares of the analysed universe are used instead, that is how many securities per currency. It is an approximation, and it is stated here so that it stays visible. Currencies with no primary source configured are reweighted onto the covered ones: assuming their rate is zero would be a stronger and almost always false claim. If no source answers and nothing is in memory, the rate is zero and the Sharpe ratio goes back to being simply return divided by swing — and the page says so rather than keeping quiet.
What the results are not
Every figure BlissPoint shows is computed on historical data and describes how those securities behaved over the period analysed. They are not a forecast, not a guarantee of future return, and they take no account of costs, fees, taxes or your personal situation. The past behaviour of a security may not repeat itself, and any investment carries the risk of losing capital.
The formulas that turn these data into numbers are in the technical note. Methodology.
info@bpblisspoint.com