Pair Trading Lab shuts down on 1 October 2026 after fifteen years, and its operator has said a new product is coming without saying what. If you built your pairs work on its database, this page is the plain comparison: what PairDesk does that the database did, what it does differently, and what it does not do at all. Whether it is a Pair Trading Lab alternative depends on which half you used: for the database half, judge for yourself below; for execution, no. No testimonials, no press logos, and nothing here claims the pairs will make money, because the self-test published with the launch build found that they cannot be ranked for that in advance.
| what you had | Pair Trading Lab | PairDesk |
|---|---|---|
| pairs | about 9.5 million pre-analysed US pairs in a database | 1.7 million pairs tested every week on about 1,900 liquid US stocks and ETFs; a liquidity-screened universe, not every listed name |
| cointegration | cointegration tests at 95% and 99% | Engle-Granger in both orientations at finite-sample critical values, with a false-discovery q-value beside every test statistic, against a null built from real series rotated out of alignment |
| hedge ratio | orthogonal regression | total least squares, plus walk-forward ratios that were fitted without seeing the period they are judged on |
| half-life, z-score | per pair | per pair; the z is walk-forward, and the half-life carries the small-sample correction |
| profit stats | per pair, backtested without slippage, fees, taxes or dividends | inside each candidate's dossier only, never as a sortable column, because no column ranks the year ahead; costs and borrow breakevens printed beside them |
| capacity | not stated | from actual dollar volume, with the leg that binds it identified per pair |
| online backtester | yes, your own rules | no; a 108-setting sweep is precomputed per candidate, but you cannot run your own rules |
| portfolio tools | portfolio organiser and backtester | a book tab that re-forms a candidate book under name and correlation caps in the browser; it is not a backtester |
| execution | PTL Trader, automated | none, and never |
| freshness | database updated on the vendor's schedule | a new build every week for paid seats; the free build is the 21 August desk, aged on purpose, and the decay table on the landing says what the aging costs |
| delivery | web application, login, premium membership | one file that opens in any browser, no account, nothing installed, nothing phoning home; a one-page brief by email |
| the honest part | a 2008 to 2009 backtest on the home page | a self-test on the front page that says accepted pairs do not beat correlation-matched rejected ones out of sample |
The complete 21 August 2026 desk: 12,000 register rows, 700 candidates with charts, the methodology inside.
OPEN THE FREE BUILD$29, $39 or $49 a month by cadence: monthly, bi-weekly or weekly desk and brief. The first ten subscribers take the weekly desk at $39 a month, locked. Firms take a five-seat license with the register export at $199 a month.
THE DESK, IN FULL