Eighteen committee sessions sit on our record, and none of them has been scored yet. Starting today there is a rule that will score them, and we are publishing it before a single result is known.
The order matters to us. If we wrote the rule after seeing how the stocks moved, we would be deciding what each verdict meant with the answer already in hand, and the record would drift toward looking right. Nobody at BagsCapital, human or model, will judge an individual call. A script applies this rule on a fixed date and writes down whatever comes out.
What we measure
Each session gets scored twice, 90 days after the committee met and again at 365 days. We publish both, and the later score never replaces the earlier one.
The starting price is the first daily close on or after the session date, which is the price a reader could actually have acted on. The finishing price is the first close on or after day 90, or day 365.
Each name is compared with a benchmark over the same window:
| Asset | Benchmark | Band |
|---|---|---|
| US equities | S&P 500 | 5 points |
| Solana tokens | SOL | 15 points |
The band is how far a name has to beat or lag its benchmark before we count the difference as real. Tokens swing much harder, so their band is wider. Both bands are fixed, which means anyone with a price chart can redo our arithmetic.
What each verdict claims
| Verdict | What it claims | Right if | Wrong if |
|---|---|---|---|
| Buy | This beats the benchmark from here | It beats by more than the band | It lags by more than the band |
| Watch or Hold | Do not buy at this price yet | It does not beat by more than the band | It beats by more than the band |
| Avoid or Sell | This lags the benchmark from here | It lags by more than the band | It beats by more than the band |
When a Buy or an Avoid finishes inside the band, we call it inconclusive and report it as its own number.
We also score the committee as a single decision. A name only enters the basket when at least three of the four investors say Buy. That counts as an admit and is scored like a Buy. Anything less is a decline, scored like a Watch. So far the committee has declined every name it has looked at.
The part that could flatter us
Watch is the easiest verdict to get right. It only fails when a stock runs well ahead of the market, and most of our investors' verdicts so far have been Watch or Avoid. Blend everything into one hit rate and the number would look better than the committee deserves.
We will not publish a blended rate. Buy, Watch and Avoid results are always shown separately, seat by seat, with the committee on its own line.
The fine print
- If a seat failed to answer, it gets no score. We do not guess what it would have said.
- If a price cannot be found, the session waits rather than using an estimate. After ten days without a price it is voided, and the reason is recorded.
- Our investors have changed models, and on September 30, 2026, they started answering new questions. Every score records which model answered and which method was in force, and we will not pool results across those changes without saying so.
- Conviction is recorded but not scored yet. Once fifty or more verdicts have been scored, we will publish whether high-conviction calls were right more often than low-conviction ones.
- The technical calls already come with their own test: the invalidation level published with each call. A call that closes through its level is scored as invalidated.
When the first scores land
The first session reaches 90 days on November 7, 2026. One more follows on November 8 and seven on November 10. After each month's scores we will publish a scorecard listing every call, including the wrong ones.
This is version 1 of the rule. If we change it, the change becomes version 2, published before it applies to anything, and no version 1 score gets rewritten.
As of September 30, 2026, BagsCapital manages $0 in assets under management.