Let’s jump right into our 2026 MLB Beat the Streak picks:

If you’re interested in how the picks have performed historically, I’ve added a new table below. Occasionally lineups differ from morning projections, so those cases are noted in grey in the Outcome column.

Anyone who’s played Beat the Streak knows the goal of the game: give you just enough hope for a miracle run once per season, then spend the rest of the year brutally crushing your dreams. I expect this year to be no different, and you should too.

However, if someone does manage to push the streak over the line, let the record show I’ve always believed in strength in numbers—and that 2026 will be our year.

Now, there’s a few different ways to present these results, but I’ll start with the topline results.1

The average predicted probability for the top 20 players each day was about 68%, and they came through about 70% of the time. However, overall averages aren’t really the point, because you probably weren’t picking the 17th-best option. So here’s a better way to view this data that shows how the top tier performed:

The table above splits the predictions made by the model across the entire season into 12 buckets. Each bucket represents a range of predicted probabilities that a batter will record at least one hit in a given game. For example, the first bucket spans 71.7% to 79.3%, meaning a player predicted at 73% would be included in that group. For each bucket, the table shows the model’s average predicted probability, the actual hit rate observed, the resulting calibration error, and the number of observations in that bucket.

At the top of the board, the model missed by just 0.7%. Overall, it was reasonably well calibrated, with most buckets showing actual outcomes slightly exceeding predictions. For this project, that bias is generally preferable, though my goal next year will be to tighten it without introducing overconfidence.

I’ve also included the top 6 bins as dot bars (is that a real name for a chart type?), with each dot representing a single model pick and its eventual outcome. The bins are labeled with their rate of correct picks. I probably did 30 iterations of this visualization before landing here, and I’ll say it’s surprisingly fun to hover over the individual dots and see the guys who helped — and ended — my streaks this year.

Because I’m interested (and some of you might be as well), I included a “BTS Legends” table featuring the top five qualifying batters by game hit rate in 2025.

Witt Jr., Bichette, and Ohtani being on this list come as no surprise to me, but I don’t believe I picked Naylor or Bohm a single time this season. I must’ve picked Bobby Witt Jr. 100 times this season, and for good reason: if you’d stuck with him alone, you could’ve beaten the streak twice with games to spare.

Maybe next season someone will beat the streak, or better yet, do it using my model. Speaking of next season, the new model isn’t done yet, and March 25th is coming fast. I hope you’ll join me again for another attempt to beat the streak in 2026 — however fruitless it may be.

That’s why you’re getting The Scorecard: a new article format I’m using to look back at how my models actually performed, focusing on accuracy, calibration, and what the results mean going forward.

Accuracy here is simply a measure of how wrong I was. So that’s where I’ll start:

Across all state and district contests, my model missed by an average of 2.43 points, with smaller errors in the states that mattered most. The eventual winner was favored in 49 of 50 states, D.C., and the Nebraska and Maine congressional districts that award an electoral vote.

A more compelling way to see this is on a map. Below are my final predicted margins from election night, followed by the actual outcome.1

The maps are similar overall, with the most obvious difference being that the actual outcome is noticeably redder. This pattern wasn’t unique to this model. Across the field, election models in 2024 tended to miss in the same direction, slightly in Democrats favor. This kind of shared miss is an expected feature of election modeling, where polling errors are highly correlated across states rather than independent, a point Nate Silver emphasizes repeatedly.

So, if everyone missed together, how did the models compare?2

Given the shared direction of the miss, the distinction becomes magnitude. And by that measure, this model produced the smallest average error across states, especially swing states, and favored the eventual winner in more contests.

That difference is easier to see here.

Each point represents a state’s predicted margin versus its actual result, with distance from the diagonal reflecting error.

What Makes This Model Different?

The biggest distinction between this model and the others is not how it interpreted the polls, but how much weight it placed on signals that had nothing to do with them. On the morning of the 2024 election, I noted how predictive fundamentals had been in the previous cycle and briefly explained what they are in my election-day breakdown. I also noted that while many probabilistic models largely phase out fundamental data as election day nears, this model would not. My view is that fundamental data should be used to the extent that it remains predictive, rather than phased out by default. That choice meant the model would look meaningfully different from the field, and could just as easily have hurt performance as helped it.

Ultimately, 2024 became another case study in how predictive fundamentals can be, even late in the cycle. In fact, the fundamentals-only model would have been the strongest performer in 2024.

Remarkably, the fundamentals-only model favored the eventual winner in every state and district without relying on a single state-level poll.

This doesn’t mean future models should be fundamentals-only. The takeaway from 2024 is that when polling errors are shared and highly correlated, incorporating a strong, independent signal like fundamentals can help smooth those errors.

Presidential elections are rare. That makes modeling decisions consequential, because if you miss the mark, you have to wait four more years to correct course. My view is that, over time, fundamentals will improve model performance more often than not. And if I had to make one final forecast, it’s that by 2028, more models will be incorporating fundamental data more fully into their final forecasts, and the novelty of this model may wear off.

Update 9/28/25: The 2025 MLB regular season has come to a close. To answer the question in the header, we did not beat the streak in 2025. The updates will resume on opening day 2026, but I’ll leave the page as is so the concept is still clear. I hope the frequent viewers out there found this useful. The plan is to release the final calibration numbers so we can judge how well this model worked. See you next year!

Let’s just start with today’s picks, I’ve put the boring stuff below.

Okay, I promised you the boring stuff so here goes:

A little bit of background

Beat the Streak is a competition launched by the MLB in 2001. The premise is simple: pick 57 batters in a row to record a hit, surpassing Joe DiMaggio’s 56 game hit streak, and win 5.6 million dollars. If you aren’t familiar with this nearly quarter-century-old challenge, you can learn more about it here. And while the premise is simple, the execution is nearly impossible.

It’s not hard to see why, either. Even if you select only the best batters with a .300 average and assume they’ll get around 4 at-bats per game, the odds of that batter recording at least one hit in a game are:

A 76% chance is a solid bet, but even the best bets turn astronomical when you need them to hit 57 times in a row.

The odds of this happening are roughly 1 in 6,300,000. To put this number in perspective, you’re about 400 times more likely to be struck by lightning in your lifetime than to beat the streak.

But…

As with all statistical problems, there’s power in numbers.

How the streak could be broken

For simplicity’s sake, let’s say our odds for each pick are 75%.

If you do the math, the odds of getting a 57-game streak in this case would be 1 in 13,228,420. That means for every 13,228,420 attempts to beat the streak, we would expect one success. And because there are 162 games in the MLB season, and Beat the Streak allows you to pick two players per day, that means 324 picks can be made in a season.

We can also figure in that every team gets days off, and they’re staggered, so there are more like 180 MLB game days in a season. We also need to note that once there are only 28 game days left, and you don’t have an active streak, you would no longer be able to reach 57, even if you played perfectly for the remainder of the season. 28 days × 2 picks per day = 56, which falls just short of the 57 needed. That means only the first 152 game days are viable for a full run.

If we take our odds of 75% per pick, we would expect 1 in every 4 picks to miss, leaving us with an average streak of only 3 (sad). But the good news is that every two days we can start our streak anew, giving us:

152 game days ÷ 2-day streak = 76 total attempts to Beat the Streak per season.

If we also factor in that some people have lives and don’t want to make picks for 152 days straight, or that they may want to take things slow and lower their pick frequency when they have a strong streak going, then it’s more realistic to expect they’ll make around 38 attempts per season.

So…

If we need 13,228,420 attempts at the streak, and each person makes 38 attempts, then we would only need:

13,228,420 ÷ 38 = 348,116

Just 348,116 people playing Beat the Streak this year would give us a reasonable expectation that someone finally pulls it off.

Is it realistic?

According to MLB, more than 5 million unique users have participated in Beat the Streak, and two users have achieved streaks above 50. So from an intuitive standpoint, I do believe it’s realistic. Even I had a pretty good streak going at 29 back in the 2019 season (I’ll never forgive Charlie Blackmon).

If there were renewed interest in Beat the Streak, and a sense of hope to keep players engaged, I think we could get someone over the finish line.
Perhaps all we need is for my massive following here on Substack to mobilize and greatly increase the number of consistent streakers we have (interesting word choice) and we could accomplish what many thought impossible.

The inspirations for this project were the probabilistic forecasts from 538, Nate Silver, and The Economist. However, my model differs from these forecasts because it also draws significant inspiration from deterministic forecasts such as Election Predictions Official and The 13 Keys to the White House. The former create probabilistic outcomes by aggregating polling data and applying uniform polling errors to create a mix of scenarios, while the latter assess different fundamental factors to either pick a winner in each state or overall. While probabilistic models do consult 'fundamental' data, it’s largely phased out as the election nears.

2020 was a great year for the fundamentals. The 13 keys were right, and Election Predictions Official correctly predicted 49/50 states.

My state fundamentals metric weights each state's past voting patterns to determine its 'lean.’ After the ‘lean’ is determined, I add it to the national popular vote estimate . For instance, if Ohio historically leans 8 points to the right of the national average, and the country is estimated to vote 2 points to the left, then Ohio would be predicted to vote for the republican by 6 points.

This metric was a stronger predictor of the final results than the polling averages in the last election cycle. The following graphs show the predicted Republican margin of victory in each state compared to the actual Republican margin (a negative value indicates a margin of defeat). The grey line represents perfect accuracy.

My forecast derives its final odds from both of these metrics (along with betting market data). However, because of this, my model is significantly more “Trumpy” than other probabilistic models. The fundamentals appear especially “Trumpy” largely because of the popular vote estimate. The popular vote estimate is a regression equation that considers the national polling average, economic data, whether the incumbent party nominee is the sitting president, and the incumbent party job approval. Of these, the national popular vote polling average carries the highest weight. I think of all the numbers and data to speculate at, the national popular vote polls have been the most concerning for Harris, even though the national popular vote does not determine the winner.

In 2016, Hillary Clinton won the popular vote by 2.1 points yet lost the electoral college by around 77,000 votes spread out between the three decisive states of Wisconsin, Pennsylvania, and Michigan. In 2020, Joe Biden more than doubled the popular vote margin to 4.5 points and won the electoral college by just 43,000 votes spread across Arizona, Georgia, and Wisconsin.

Harris leads by just 1.1 points in The Breakdown Point’s final national polling average and an even slimmer 0.9 points in the forecasted popular vote. Other models show similar results: 538 has Harris up by 1.2 points, Nate Silver by 1 point, and Real Clear Politics by just 0.1 point.

Because my forecast makes a deliberate use of the national polls to inform the fundamental data, Trumps odds have been great since October 1st.

My opinion is that a forecast like this is a good way to think about elections, especially in the Trump era. I got further confirmation of this when I compared my results to Polymarket, the world’s largest betting market.

My final numbers are still better for Trump than Polymarket, and perhaps this is due to the lack of conversation around the national popular vote and the current political environment appearing to be much better for Trump than 4 years ago. Maybe Polymarket’s odds are weighed down by the forecasts of Silver, Lichtman and others. Or more likely, it’s that those models are better representations of the reality on the ground. If the polls outperform the fundamentals, then this thing truly is a tossup. If the fundamentals outperform the polls yet again, then Trump probably walks away with over 300 electoral votes. To simplify my forecast, it weights these two scenarios as roughly equal, giving a slight edge to polls.

As for me, I really don’t know what will happen tonight. I think I land pretty close to what the betting market thinks. My gut says Trump will win with about 60% certainty, but to put this in perspective, let me give my gut’s record:

• 2008: Obama, 95% certain

• 2012: Obama, 75% certain

• 2016: Clinton, 99% certain

• 2020: Biden, 80% certain

Not terrible, although the year I was most certain was the year I was most wrong. It has been an interesting election to say the least, and tonight should be interesting as well.

If the model does well enough, maybe there will be a recap article after the election. If it does poorly, then I’ll probably bury my head in shame. Either way, I hope you come back next year as I plan to get my baseball data projects out before the 2025 MLB season kicks off next March.

FINAL UPDATE: Tuesday, 11/5/24 at 7:00am PT

Note 11/5/24: The “tossup” category has been removed for the final state-level predictions.

Let’s start with a look at the 2024 Electoral College Map. Because what’s a good election forecast without an electoral college map?

The electoral college map reflects the current state of the race. The model establishes a baseline margin for each state by using economic indicators and past voting patterns. It then adjusts this baseline with up-to-date state polling and betting market data to capture current trends in the political environment.

Now let’s take a look at an alternative depiction of the electoral college map. Because two electoral college maps are better than one.

This map scales each state according to its electoral college share. Each tile represents 1 electoral vote.

What goes into these predictions?

Great question, the following table gives some insight into the metrics used to predict the final margin of the national popular vote and 16 most competitive state contests.

So what does all this mean? Who is going to win?

That really is the 3,267,039,019 dollar question1. And I don’t have the answer. What I do have, however, is another chart. Below is the outcome of 44,000 simulations of the 2024 election and the results summarized in the final 2024 election top line:

If you’re interested in how the horse race has changed over time, I have one final visualization showing the trend lines for each candidate since I published the forecast:

My understanding of this race, after spending hours poring over data, preparing thousands of lines of code, and staying up until 3 a.m. because I “just can’t get this thing to work,” is fundamentally unchanged from where I started. I think this race is a toss-up2, and I wouldn’t be surprised if either candidate walked away with over 300 electoral votes. The only thing that would surprise me is if the election comes and goes without controversy.