Extra Sensory Perception Network (ESPN)

ESPN predictions tend to be one of the biggest factors in decisions regarding FA pickups or setting up the active rosters week after week. A lot of times we are very glad and have no problem when our player performs better than ESPN predicted. But if a lot of our players do not play as well as ESPN predicted, it is often a cause for a lot of heartburn. So Vishal's question is -- just how good are ESPN predictions? Mind you that the data I have is for 3 weeks only. I will revisit this topic again when the season is ending.

3 weeks, 12 teams per week, 14 players per team ... that amounts to 504 players so far on our rosters.

First question - Are ESPN predictions on an average aggressive or conservative? Do they overpredict numbers or underpredict them? What do you think?

Out of the 504, 242 players outperformed or equalled ESPN's predictions. Mind you that also takes into account 26 players who ESPN predicted would not score any points and did not. Some of them were good calls by ESPN I am sure, but a lot of others were either injured or werent playing ... so it is not in that sense a prediction at all if we all knew that they were out of the lineup and would score nothing.

Let us take those 26 players out of the equation --
Out of 478 players, 216 players outscored or equalled ESPN predictions. 262 did not. So 55% of the players underperformed and 45% were consistent or better than ESPN predicted.

The good thing about the stats above is that ESPN neither overpredicts nor underpredicts. Short of being able to tell the future, a 50% over and a 50% lesser is the best result that you hope to achieve. ESPN is close to that with a slight tendency to overpredict.
The bad thing about the stats above is that since it does not tend to overpredict or underpredict, it does not tell us anything more about how to use ESPN's predictions better.

Apart from the number of players going over and under, another important yardstick is by how much? And that would be in my mind more important. Because if ESPN predicted 2 players would score 10 points, and 1 of them scored 9 and the other one scored 11, then it is all good. However if one of them scored 2 and the other one scored 18, then obviously they are doing a terrible job.

Let's look at it in a different way. Again, I am NOT taking into account the 26 players who ESPN predicted would score 0 and did. This brings their
reliability down by a lot... keep that in mind while you soak up these numbers:

Out of the 478 players, 24 players scored exactly equal to what ESPN predicted -- 5%
, 77 players scored within 1 point of what ESPN predicted -- 16%
, 118 players scored within 2 points of what ESPN predicted -- 25%
, 227 players scored within 5 points of what ESPN predicted -- 48%
, 299 players scored within 7 points of what ESPN predicted -- 63%

This means that there is around two-thirds of the players will score around 7 points of what ESPN predicted.
Around half of them will score within the range of plus or minus 5.
And a quarter of them will score within 2 points of their predictions.

But we dont care when our players overperform right? So we just need to know what are the chances they will perform within acceptable limits:

FINE!

Out of 478 players, 216 players scored equal or higher than their predictions -- 45%
, 299 players scored not less than 3 points than their predictions -- 63%

This means that when you pick a player to play in a week, there is a 2/3 chance of him or her scoring within a 3 point grace for you. That cannot be bad.
Also when you pick a player, there is almost a 50% chance he will do as well or better than predicted.

Looking at these numbers, ESPN does do a decent job. They arent bang on most of the times but to me, these are acceptable limits. Let's be honest -- they are trying to predict the future here.

NOW LET'S GET TO THE FUN STUFF

ESPN-SCREWED-ME-AWARD goes to
NIRAL JHAVERI ... On an average per week, Niral's team's total score is 38 points less than what ESPN predicts!!
Close second is Valmik Desai ... 36.33 points less!
Both teams have a 0-3 start. That makes sense. However you'd never guess who was third behind Niral & Valmik on the espn-screwed-me list. A 3-0 guy! Saumil Acharya ... even though his team scores almost 30 points less than advertised, he still has won all 3 games ... KUDOS! Saumil has had a very lucky start to his season for sure.

I-KNOW-MORE-THAN-ESPN-AWARD goes to (only considering active players not benched players)
Ali & Shilpi/Jimit
They are the only ones who have outscored their ESPN predictions overall.

No one team has outscored their ESPN predictions all 3 weeks!

MY-PLAYER-LOVES-ME-BACK Award goes to Shilpi & Jimit -- Peyton Manning has scored 90 points for them this season. More than any other player for any other team. Close second is Drew Brees for Devashree with 87 points.

BTW ... this does seem like the YEAR OF THE QB ... Top 4 point scorers so far have been QBs. 7 out of the top 10 scorers this far have been QBs! 13 out of the top 20 --QBs!!

ESPN doesn't love me award goes to:
Giants D/ST
Nate Kaeding
Nick Folk
Donald Brown
Bears D/ST
Cedric Benson
Ben Roethlisberger
Jets D/ST
Stephen Gostkowski
Percy Harvin
Patriots D/ST

The 11 players below have consistently outscored ESPN predictions in all 3 weeks.

On the other end of the spectrum, there is no player who underperformed espn's predictions all 3 weeks!

And finally, this is something I want to do every week ... in our format of FF, we play head to head. What if everyone played against everyone else every week. What would the win-loss record be?

The standings after week 3:
Team Win-Loss-Tie
-----------------------
Murtaza 24-9-0
ShilpiJimit 22-11-0
Devashree 20-13-0
Meghna 19-13-1
Prashant 19-14-0
Saumil 18-14-1
Darshini 18-15-0
Vishal 16-17-0
Shirin 14-18-1
Valmik 9-23-1
Niral 8-24-1
Ali 8-24-1

Saumil is 6th and still 3-0!!! LOL! That's called good matchups and luck! ;)

Phew! That is it boys and girls!

Remember the BYE weeks start from this week so make sure to keep an eye out for those players on BYE.

Happy Sunday!


P.S.: Thanks to Vishal for this suggestion of dissecting ESPN's predictions. If any of you have any ideas, let me know and thy shall be done!


COMMENT ON COMMENTS:
Thanks Vishal, Shirin, Valmik & Saumil.

Valmik -- Point taken. I know you used exxageration because no one player will ever score or be predicted to score 200 points. However when I looked at the trends for predictions v/s actuals of players predicted to score above 15 and then predictions v/s actuals of those predicted to score less than 15, the trends were quite similar and hence the overall trend was the same too. That is why I omitted that part. Hopefully in the second half when I revisit this, I will definitely look at that too.