A Winning Streak Is Worth Twice as Much in May as in April. And Nothing at All in August.
Boston won 14 in a row in July and it saved their season. If the Mets won 14 in a row tomorrow, their playoff odds would move by 1.6 points. Same streak. Different month.
The question: when is a hot streak worth the most?
The answer: the middle of the season. A 14-game winning streak would have bought the Mets 26 points of playoff probability on May 18. Today it buys 1.6.
Here is the surprising part. Earlier is not better. On April 20 that same streak was worth only 12.5 points — less than half of May’s. Get hot too early and it washes out.
What you get from it: two questions to ask the next time somebody tells you a team is red hot.
How we put a price on a streak
Pick a date. Take the real standings that morning and the real remaining schedule for all thirty clubs. Play the rest of the season 20,000 times and count how often the Mets reach October.
Now do it again, and make the Mets win their next fourteen. Count again. The gap between the two numbers is what the streak was worth on that date. That gap is a counterfactuali, and it is the only honest way to price a run of wins.
Too early is a real problem
Look at the first two points on that line. They are almost a controlled experiment.
On April 20 the Mets were six games out with 140 to play. On May 18 they were six and a half out — slightly worse — with 115 to play. Same hole, twenty-five fewer games, and the streak is worth twice as much.
The reason is simple. A streak ends, and then you go back to being whatever you actually are. The Mets are about a .470 club. Fourteen wins in April get followed by 126 games of being a .470 club; that is plenty of time to erase them. In May only 101 games remain to do the erasing.
That is regression to the meani wearing a uniform. The season is long enough to take your hot streak back.
A run of wins is not a gift. It is a loan against the games you have left.
The ProfessorToo late is the obvious problem, and it arrived fast
Now find the cliff in the middle of the chart. On June 22 the streak was worth 22 points. One week later it was worth 3.
Nothing about the calendar explains a drop that steep. The Mets went 3–10 in those seven days and fell from six games back to ten.
So be careful with the lesson. Time was not the only thing they lost; they also lost the closeness. A streak needs both. There has to be enough season left to bank it, and you have to be near enough for it to reach.
Miss either one and fourteen straight wins buy you almost nothing.
Boston had both, and cashed the ticket
On July 2 the Red Sox were 37–48. They were six and a half games out of the last wild card with 77 to play, and our model gave them a 5.3% chance.
Then they won fourteen in a row, July 3 through July 20. By the end of that streak they were at 69%. Today they are at 97.5%.
They did it while missing Garrett Crochet, Trevor Story, Triston Casas and Roman Anthony. Remember that the next time a lost season gets blamed on the injury list.
What the Mets would need, and whether it has ever happened
On the morning of August 22 the Mets were 58–71 with 33 to play, eleven games out, and tenth in a race for three spots. They have seven teams to pass.
To finish at .520 they must go 27–6. That would still leave them only a one-in-four shot. To be a genuine coin flip for October they need 28–5.
We checked every game the franchise has played since 1962 — 10,255 of them, which contain 8,175 different 33-game stretches. Exactly one of them is 27–6: June 4 to July 12, 1990.
Hold onto that one. It is the best 33-game run in the history of the New York Mets, and that team finished 91–71 and missed the playoffs.
This year the Mets’ longest winning streak is four games. They have done it four times and have never won a fifth. Boston’s best 33-game stretch this season is 28–5 — better than anything the Mets have managed in sixty-five years.
What to take home: two questions
“They’re red hot” is not information. It is a mood. Turn it into information with two questions: how far back, and how many games left?
Both have to come out right. Close but out of time is the Mets in August; plenty of time but nowhere near is a bad team in April. Boston in July was the rare case where both were true at once, and that is why one good fortnight changed everything.
Take the pair of questions with you. Think of a sales quarter that turns around in week eleven, a poll bounce in October, a fund that has a wonderful March. Each is a run of good results; each is worth only as much as the time left to bank it.
Watch for the April version too. A hot start feels like proof, and it is the least valuable heat of the year.
Notes & sources
Standings, schedules and results from the MLB Stats API, current through the morning of August 22, 2026. Every team’s games played plus games remaining was checked to equal 162 before any simulation was run; postponed games that had not been rescheduled were excluded, which is what made the first pass fail that check.
Method: each team’s strength is a 50/50 blend of its actual winning percentage and its Pythagorean expectation from runs scored and allowed, regressed toward .500 by 69 games. Game probabilities use log5 with a home-field adjustment set at a .535 league home win rate. Every remaining game in both leagues is simulated, division winners and three wild cards per league are taken from the final standings, and ties are broken at random rather than by the real tiebreaker rules. 100,000 seasons for the headline odds, 20,000 for each point in Figure 1.
The result is not sensitive to the regression choice. With no regression at all the Mets come out at 0.001% and Boston at 98.6%; with heavy regression, 0.007% and 97.0%. The figures quoted above use the middle setting.
Franchise history covers all 10,255 regular-season games the Mets have played from 1962 through August 22, 2026, taken game by game rather than from a summary. Stretches are counted within seasons, never across an off-season, giving 8,175 windows of 33 games. Two overlapping windows reach 27–6, both inside the same June–July 1990 run.
One limit worth stating. Forcing a team to win fourteen games does not make it a better team in the simulation; it simply hands over the results. A real fourteen-game streak might carry information we are throwing away — a returning starter, a fixed bullpen. If so, these numbers understate what a streak is worth, and they still do not save the season.
