Do Hot Streaks Predict Home Runs? What 162,494 MLB Starts Say

162,494 starts by MLB batters from April 2023 through September 25, 2026 — regular season only, starters only, and only once a player has 15 games behind him so there is a season to compare against. A start counts as a hit if the batter homered at least once. Across all of them, that happened 11.9% of the time.

One cleaning step mattered more than any other: postponed and suspended games appear twice in the raw box score feed, under the scheduled date and the played date. Left in, a player's “last five games” can include tonight's result, and every streak looks predictive. We keep one row per player per game, dated to when it was actually played.

The comparison that matters is not “hot hitters versus everyone” — hot hitters are mostly good hitters, so they beat the league average without the streak adding anything. The fair comparison is hot hitters versus hitters with the same season-to-date power, measured only from games before the one being predicted.

Home run rate in the next game, by homers in the previous five:

Read across the bottom row. A hitter with three or more homers in his last five went deep in 17.4% of his next starts. A hitter with the same season-to-date power and no streak: 16.2%. The streak is worth about a point — real, but a small fraction of the gap the raw numbers suggest, because most of a “hot” hitter's edge is simply that he is a power hitter.

Homering the night before is the same story: 14.2% the next day against 13.8% for hitters of the same power. The popular “hot streak” definition built on batting average (hitting .300 with power over five games) is worse still — it can miss low-average sluggers entirely. Pete Alonso homered his way through the last week of September 2026 without ever tripping it, so it measures the wrong thing.

When we added recent-power inputs (homers and extra-base hits over the last ten games, a weighted recent home run rate, a power-hot flag) to a home run model on its own, it did not reliably rank the next night's hitters better on seasons it had not seen. That matches the table: the streak is already mostly inside the season numbers.

The mirror image is the slump. Restricted to the top tenth of hitters by season-to-date power:

A slugger in a 15-game drought homered in 18.6% of his next starts — the same as sluggers with one, two or three homers in that stretch. The drought sample is small (236 starts, because elite power hitters rarely go 15 games without one), but there is no sign that a cold slugger is a worse bet than his season says. Even the best of them go homerless about three games in four: in 2026, the highest rate among regulars was Munetaka Murakami's 26.4% of starts.

The other half of every prop write-up is the opposing starter. We sorted every start into fifths twice — once by the batter's season-to-date home run rate, once by the starting pitcher's season-to-date home runs allowed per batter faced (both shrunk toward league average early in the season, both using only prior games):

The batter column runs from 6.9% to 17.4%, a 2.5× spread. The pitcher column runs from 10.9% to 12.7%. The pitcher matters — the most homer-prone starters give up about 17% more homer games than the stingiest — but a strong power hitter facing the league's best home run preventer is still a better bet than a weak hitter facing its worst. Part of the reason is structural: the starter faces a hitter two or three times, and the bullpen gets the rest.

Where the two meet, the effect is additive and modest: the top-power fifth ran from 16.1% against the stingiest starters to 18.7% against the most generous. Worth a tiebreak between two similar hitters. Not worth building a card around.

“He's taken this guy deep before” is the most quoted stat on a home run prop. Hitters who had homered off tonight's starter at some point went deep 13.4% of the time — but their own season rates said 14.1%. Hitters who had faced the starter without homering: 12.0% against 11.8% expected. Batter-vs-pitcher samples are ten or twenty plate appearances; the hitter's power over hundreds of plate appearances swamps them.

It is also the easiest stat to fool yourself with. Before we removed the duplicated box score rows, head-to-head history was the single strongest input in our test model — because a duplicate row let the “history” contain tonight's homer. Clean the data and the signal disappears. If a stat looks too predictive, check whether it can see the answer.

• Start from the hitter's power, not his week. Season contact quality and home run rate carry almost all of the information; the last five games add about a point.
• Don't pay up for a streak. Books and bettors both react to recent homers. If the price moved because he homered twice this week, the move is bigger than the signal.
• Don't fade a cold slugger. A drought is not information about the next game.
• Use the pitcher as a tiebreak and ignore head-to-head history. Park and weather deserve more attention than either — see what weather actually does to home runs.
• The real question is whether his power changed. That is a different problem from a streak, and it takes longer to see — how many plate appearances before a home run breakout is real.

Even done well, this is a low-probability market: in these four seasons a starter homered in about one game in eight. We publish our top three home run picks every day before first pitch and grade them in public, so the record is there to check rather than to take on faith.