How We Filter Our Picks
We built a multi-variable model that calculates a true line for every MLB game. It was profitable from day one.
Then we nearly doubled its ROI without changing a single variable or weight — just by filtering which picks we actually recommend.
The discovery
After tracking hundreds of MLB games, we ran an exhaustive analysis — testing every combination of edge thresholds, moneyline ranges, home/away splits, and category restrictions we could find.
The data told a clear story: home picks are profitable at lower edge thresholds than away picks. Road games carry more variance that our model can't fully capture — travel fatigue, unfamiliar parks, hostile crowds, and bullpen matchup uncertainty all add noise. Home picks convert smaller edges into profit because the home environment reduces that variance.
Why home and away need different thresholds
Consider two picks with the same edge. The home pick benefits from the team's home environment — the model's projection is more reliable because it's working with a more predictable situation. The away pick carries extra uncertainty that the model doesn't explicitly account for.
Our data across hundreds of games confirmed this pattern decisively. Home picks were consistently profitable at a lower edge threshold. Away picks needed a significantly higher edge before the signal overcame the noise.
How the filter works
The rules are simple:
Home picks are included above a lower edge threshold. The home environment makes smaller edges profitable.
Away picks are included only above a higher edge threshold. Away games need a larger edge to overcome the extra variance.
That's it. No moneyline thresholds, no favorite/underdog distinctions, no complex carve-outs. Just two proprietary thresholds — one for home, one for away — derived from exhaustive analysis of our data.
Everything else is filtered out and labeled No Play. The game is tracked in our database for analysis, but it's not recommended to users.
The results
The picks that were filtered out were net losers collectively with a negative ROI. Removing them concentrated the output on the model's actual strengths.
Best Bets vs Undervalued
The filter naturally creates two tiers of picks.
Best Bets are picks where the model calculates an edge above our highest-conviction threshold, regardless of side. These are the strongest mispricings we find — typically two or three per day. When the model is this confident, sizing up makes sense.
Undervalued picks are everything else that passes the filter — included picks that clear the home or away threshold but fall below the Best Bet line. These are the volume plays. Individually, any one might lose. Collectively, they produce consistent returns.
No Play is everything the filter removes. You can still see these games on our platform if you choose “Show Everything,” but they're dimmed and clearly labeled. We track them for transparency and ongoing research.
Why simplicity wins
We tested dozens of more complex filter rules — moneyline caps, favorite/underdog distinctions, multi-range edge carve-outs. The complex rules performed similarly or worse than the simple two-threshold approach, and they were more likely to be overfit to the specific games in our sample.
Simple rules generalize better. Two thresholds — one for home, one for away — are easy to verify and hard to overfit. They capture the one structural truth our data consistently shows: home games convert smaller edges into profit.
See the filter in action
Every pick on Dr. TrueLine is labeled Best Bet, Undervalued, or No Play. You can filter by tier, check our track record for each tier separately, and verify that the filter produces exactly the results described here.
We don't hide the No Play games. We don't pretend every game has an edge. Transparency isn't just our policy — it's our best sales tool.