A Quick Note On Predictions
Opening Day is Monday, so over the next few days, you’re going to be bombarded with season previews full of predictions. Most of them will have the Mariners finishing 3rd or 4th, somewhere between 70 and 80 wins. That’s the general consensus of most baseball observers, anyway.
You know what my prediction is? That eventually, we’ll all realize that predictions are rather pointless, and the 2009 Mariners are one of the obvious examples of why.
I think most of us reading this probably agree that this is a ~78 win team on talent level, as it stands right now. If you hate defense and are all about hitting, maybe you think it’s a ~75 win team. If you think Junior’s going to hit the juvenation machine and the bullpen’s going to sort itself out, you might think it’s an ~81 win team. But, we’re all kind of hovering around that 78 win average. So, then, we should all predict that the M’s will win 78-ish games this year, right?
Except there are about 4,000 variables that will play a significant factor in how the team does that no one has absolutely any way of predicting. If Ichiro’s fatigue turns out to be some kind of career-ending virus, the team just lost four or five wins. Are any of us knowledgeable enough to think that we have any insight into what is actually wrong with Ichiro? Of course not.
So, now, we have this significant variable that we have no insight on. If it turns out Ichiro is sidelined for a significant period of time, the team’s chances of contention drop dramatically, and the chain reaction leads to a significant increase in likelyhood of Beltre and Bedard being traded. But will they be traded in May or July? That matters, as two extra months of those guys in uniform is worth another couple of wins. Do any of us have any insight on when another team might become in need of a third basemen and offer the Mariners a package they can’t turn down? Of course not. We have no way of knowing if Team X is going to lose Player Y in May or June or July, but the timing of that injury could have a real effect on the roster the M’s put on the field.
Now, all of the sudden, the chance of Ichiro missing time has cost the team six to eight wins, thanks to the chain reaction.
Let’s play this the other way, just for fun. Say Ichiro’s fine, and the team starts the season out well, taking advantage of the injury problems the Angels and A’s are having. So now they’re in first place in mid-May, and while Russ Branyan is doing okay, Nick Johnson’s available and healthy. The M’s make a move to upgrade their first base spot, and all of the sudden, they’ve picked up another couple of wins. They continue to play well, they don’t trade Beltre or Bedard, and they’re adding pieces at the deadline. Now, maybe we see them as an 83 win team, thanks to the upgrades they make.
Just based on the resulting events of Ichiro’s health report and the actions it could spawn, we might be looking at +/- 10 games in the standings. And we’ve already said that none of us have any reason to believe that our prediction about Ichiro’s health is meaningful in any way. Any prediction we’d make would simply be a guess, because we just don’t have any real information.
So then, what’s the point of a prediction, if the entire premise on which stands can be unraveled by variables that we have no way of accounting for? The writers who predicted the Mariners would be contenders last year, for instance, obviously weren’t counting on the team giving thousands of at-bats to guys who started the season in Triple-A. The team that finished 61-101 isn’t the same one that started the 2008 season. How on earth could Baker and Stone and all the ESPN guys have incorporated into their prediction that the team would end the year with Jose Lopez playing first base in order to get a longer look at Luis Valbuena?
There are just so many things that can’t be known in April that matter throughout the year that predictions are basically worthless. There’s no point in pretending that we can know what other people are going to decide to do four months from now, and when those decisions have major impact on how the team will perform throughout the year, we have to admit that our ability to predict a final outcome is no more than a guess.
What we can do is estimate probability, based on what we know today. We’re pretty sure that the Rays are better than Royals, and we can make statements about likelyhoods based on today’s facts. But those facts are going to change, and we’d be foolish to not change the likelyhoods as they go. Most people would call that waffling on a prediction – I’d call it actual analysis.
So, how many games are the Mariners going to win this year? I have no idea. I think the team, as currently composed, is about a 78 win team, give or take five wins either way for normal amounts of luck. But if the Mariners trade Adrian Beltre in two weeks because the Cardinals find out that Troy Glaus is actually out for the year, I’m going to revise my estimate based on new knowledge.
Predictions are useless. I’d rather be in the analysis business.
Comments
12 Responses to “A Quick Note On Predictions”
Leave a Reply
You must be logged in to post a comment.
Yeah, but isn’t that part of the reason we analyze stuff? So we can find out what’s predictive?
78 seems about right to me too.
Sure – but we should be honest about what we can predict. And there are so many things that can change as the year goes on that trying to predict how this team will do in August/September is basically a fool’s errand.
Huh. So there is a reason for why they play the games after all! Who knew?
jk/ 😉
Nice piece, predication is addictive so it is good to occasionally have a voice of reason throw a bucket of cold water on the process.
Proof that predictions are useless. The Royals can’t contend with Horacio in their rotation. That’s pathetic.
That sounds like a false dichotomy to me. Predictions require analysis and a comprehensive analysis of an organization’s talent naturally leads to a prediction of how many games they are likely to win.
Obviously some things will happen during the course of the season that we won’t expect, but that’s why we acknowledge that there’s a variance around the mean.
But if the Mariners trade Adrian Beltre in two weeks because the Cardinals find out that Troy Glaus is actually out for the year, I’m going to revise my estimate based on new knowledge.
Of course you would revise your estimate given new knowledge, but just because something unexpected happens doesn’t make your initial prediction wrong.
Now, if you made a lot of these predictions and were consistently seeing the results differ from the predictions by more than your predicted variance, then you ought to adjust your predicted variance. Personally, I think that a 6-8-win standard deviation around the win total is probably more appropriate than 5 wins. In ’07, the good prediction systems were off by about 5 wins per team and in ’08 the good prediction systems were off by about 10 wins per team, but I suspect those values are pretty close to the extremes.
Honestly guys– I think we could win it all this year, we’ve got all the parts in place… This is the year we… (Oh wait, that was last year, and look where those predictions got everybody– Nearly everyone had us winning the west or competing for it down to the wire… Plus, let’s look back at the predictions for the Rays last year?).
I think last year proves your point Dave, on two major levels.
Thanks for your analysis, though. I thought it was pretty spot on– and I think the high mark for the best-case-scenario imaginable is probably 85 wins this season (because I think the West is a little weaker this year).
Nearly everyone had us winning the west or competing for it down to the wire… Plus, let’s look back at the predictions for the Rays last year?
You’ll actually find that around these parts the Mariners were expected to be mediocre and the Rays were expected to be an excellent team.
Nearly everyone had us winning the west or competing for it down to the wire… Plus, let’s look back at the predictions for the Rays last year?
ESPN/AP aren’t the best place to turn to when it comes to analysis of teams they got great reporting their analysis style might be a little behind those crazy SABR people on the internet though.
The Mariners had about six IFs last year of which at least four of them needed to happen for them to move forward and out of them I would say only one happened:
If Jose Lopez improves.
If Yuni Betancourt improves.
If Sexson is pre2008 version.
If Wilkerson produces close to Jose Guillen numbers.
If Bedard is the real thing.
If Silva is the real thing.
The list is longer this year but the hill top closer!
If Bedard is the real thing.
If Branyan can provide consistent power.
If Johjima is pre2008 version.
If Griffey and Sweeney actually provide production and “team” results.
If Silva can just keep from hurting the team most of the time.
If Wak actually gets this team especially Betancourt to take MORE pitches.
If the closer position can be filled sooner than later.
If Ichiro returns healthy and on fire to chase down those lost chances at a 200 hits season.
Give me 5 of those and 82 wins is not that hard to see.
I think that’s one common folly in a lot of preseason sports analysis: The belief that any degree of research and analysis gives you the material to make specific macro-predictions about how a season will go. There are so many variables that the best one can do is predict a general path.
I think one of the most difficult things to quantify is “team chemistry”– which is arguably a myth, and a way to rationalize unexpectedly good teams. but it would be interesting to give it a try? to come up with a statistical “characteristic” for certain types of players and them “sum” them for a team, and compare the sum to the season’s results (of course you’d have to then weight that versus the sums of the teams they competed against).
maybe go back and look at last year’s Rays, the Dodgers back in 2004, the M’s in 2008 (for the reverse effect), etc.
someone smarter than me would have to come up with the stat lines to consider (coachability? clubhouse effect? a modern dilemma, like communication- ie, number of different languages in the clubhouse… the manager himself would have an effect- almost like a field does…)
anyone ever bother to study this? i think it would be interesting to see if you can quantify it. and then interesting to compare how these numbers related to performance stats (would a “high” chemistry team outperform a high performance team on average?).
or is this just sticking a hand in a hornet’s nest?
Many people have studied it. None have found anything. If you can manage to figure it out, you’ll be a very rich man.