Quant Portfolio

For every ticker, we search through thousands of trading rules, things like "buy when the price breaks above its 50-day average and the relative strength is rising" or "sell when momentum stalls and volume dries up." We don't hand-pick the rule, the system tries them all and keeps whichever one made the most money on that specific stock over the past several years after costs.

The winning rule for each ticker is called its champion strategy. A stock with no champion that beats buy-and-hold never enters the pool.

A strategy can look great in hindsight just by being lucky, so we hold back the most recent ~1 year of data, train the strategy on the older history, then check what it would have done on the data it never saw. To earn a cup, the strategy has to beat two things on that held-out year:

• Beat the stock itself (better than just buying and holding it). Otherwise why bother trading?
• Beat SPY (the S&P 500). Otherwise you should just own the index.

Cup 4 is the strongest tier, the strategy beat both by a meaningful margin. Cup 3 beat them by a smaller margin. Anything below cup 3 doesn't make it into this basket.

If you flip enough coins, some come up heads ten times in a row. The same is true for trading strategies, run enough of them and a few will look brilliant by pure chance. To filter those out, every champion strategy gets a credibility score built from four independent checks:

• Monte Carlo significance, we shuffle the trade dates thousands of times to see how often a random reorder produces results this good. If random versions of the strategy rarely match it, that's a good sign it's real.
• Deflated Sharpe, the standard risk-adjusted return number, but penalized for how many strategies we tested. The more strategies you try, the higher the bar to be called real.
• Robustness, small tweaks to the strategy's parameters shouldn't flip it from a winner to a loser. If they do, the strategy is fragile and gets penalized.
• Walk-forward stability, we slide the training window forward through time and re-validate. A strategy that only works in one specific window is a curve fit, not an edge.

We combine those four into one credibility number from 1 (low) to 3 (high). Anything that flunks here drops out.

The survivors get sorted by a composite score that rewards both quality (cup tier) and credibility. Higher cup + higher credibility wins. Ties are broken by recent live performance.

The top of that ranked list goes through one more filter, the rules of the profile you picked above. For long only: Ranks reviewed long champions by the cup composite and sizes them proportionally to conviction.

That's why the profiles can show different stocks even though they all start from the same pool, the small-account profile drops anything under $10/share and caps the basket at 15 names, long-only drops the shorts, etc.

The cup tier measures "was this strategy historically good?" The signal chip (BUY / HOLD / OBSERVE / SELL) measures "what is the strategy doing today?" Those answer different questions, so a stock can be a cup-4 strategy AND currently in SELL, the rule did great historically, but right now it's flat or exited. We show both so you see the full picture instead of just one snapshot.

Each name uses the exact signed weight from the reviewed target publication. We do not replace those magnitudes with equal weights. Signed and leveraged targets remain visible but fail closed in simulation until borrow, restricted proceeds, and margin are modeled.