Backtest
Over this window the tool BEAT just holding SPY by 9.4 points. Active trading earned its keep here.
Model +38.5% · SPY +29.2% · EqualWeight +29.0% | BEAT SPY by 9.4pts
Equity curves
Model account from $4997.00 to $6927.68 across 352 trading days.
How the three accounts compare
| Account | Total return | Max drawdown | Trades | Slippage paid |
|---|---|---|---|---|
| Model | +38.5% | -12.9% | 420 | $135.47 |
| SPY buy-hold | +29.2% | -19.0% | 1 | $1.50 |
| Equal-weight | +29.0% | -15.1% | 15 | $2.86 |
The model made 420 trades and paid $135.47 in slippage — about $133.97 more than buy-and-hold SPY (which traded once). Frequent trading is not free; every rebalance pays the spread, so the active account starts each comparison slightly behind.
Read this before you trust the number
Kronos was pre-trained on historical market data; a backtest over an era the model may have seen is optimistic. Recent windows are the most trustworthy. Costs (slippage) ARE modeled, so frequent trading is penalized realistically.
In plain English: Kronos was pre-trained on past market data, so a backtest over an era it may have already “seen” tends to look rosier than live trading ever will. Treat this as suggestive, not proof. The real test is the forward, live run on the Portfolio tab — that one the model has never seen.
This backtest also used only 10 forecast samples per decision (lower than the live run, for speed), so the picks are noisier than production. History builds as real daily runs accumulate.
What it actually traded
| Rebalance | Picks | Evaluated |
|---|---|---|
| 2026-05-28 | TLTGLDXLVXLFJPM | 15 |
| 2026-05-20 | TLTGLDXLVXLFJPM | 15 |
| 2026-05-13 | TLTXLVXLFMSFTJPM | 15 |
| 2026-05-06 | TLTXLVXLFMSFTJPM | 15 |
| 2026-04-29 | TLTXLVXLFAAPLUNH | 15 |
| 2026-04-22 | TLTXLVXLFAAPLJPM | 15 |
| 2026-04-15 | TLTXLVXLFMSFTUNH | 15 |
| 2026-04-08 | TLTXLVMSFTNVDAUNH | 15 |
| 2026-03-31 | XLVMSFTNVDAJPMUNH | 15 |
| 2026-03-24 | XLVAAPLMSFTNVDAUNH | 15 |
Showing the last 10 of 71 rebalances. Weeks with no pick mean nothing beat the basket convincingly — so the account held cash rather than force a trade.