Tangotiger Blog

Sunday, January 05, 2020

Fantasy League Valuations

?This is how I did it 30+ years ago.  The last time I played was 30+ years ago.  I had no interest to do this blog post other than someone asked me. I have an hour, so here we are. 

If someone wants to argue that my method is lacking, feel free to point out where and why.  My only argument is: I know what I'm doing. At least, I think I do, or I've convinced myself of it. If that's not good enough for you, that's fine!  Feel free to stop reading right now.  If my argument is good enough for you, then jump over the line.  

Read More

Thursday, March 31, 2016

How would you create the stat line for a player in a board game?

I sent this to Bill James, and I'd like to hear from the Straight Arrow readers as well.

If you created a baseball board game, and you produced a card for the 1998 Shane Spencer (10 HR, 73 PA), would you set his card so he'd AVERAGE 10 HR per 73 PA?  Which means half the time, your sims would have him hit more than 10 HR per 73 PA?  Or would you set his card so that 10/73 would happen say 10 or 15% of the time (meaning setting his card at around 6/73)?  Similarly, Barry Bonds had 5 HR in 52 PA in 2005.  Would you look at his neighboring seasons (rate of around 3.5/52), see that 5/52 is not that out of line, and set his card to 5/52 (or something close)? 

Wednesday, November 18, 2015

How do you win at Daily Games?

?Off the top of my head, and having never played these games, but I can take a good guess...

There are two main ingredients:

1. knowing the starting lineups

2. knowing the ownership rate of all the athletes

More tepid ingredients:

3. a forecasting system that is better than Marcel (at least for rookies)

4. knowing the matchups (who's playing against who, and where)

From there, you write a simulator.  

5. You come up with a distribution of points for each player (which undoubtedly you can create a function for, at least one function for nonpitchers and one for pitchers).

6. You create an algorithm where you can, and you do a brute force (running a thousand sims of various combinations) where you can't

This is the kind of thing I did when I ran The Forecaster's Challenge a few years ago.  I presume there are Daily Sites out there that already do this for us?  Feel free to post their links in this thread.

Anything else obvious, and less obvious, that I missed?

Monday, November 16, 2015

How random do you make a random simulator to distinguish between luck and skill?

?If you were given the rules of blackjack, and you apply math, your chances of winning any hand is close to 50/50.  On the other hand, if you randomly hit and stick, your chance of winning any hand would be... I dunno... 20%?  So, how you decide if blackjack is a game of skill or chance depends on how much knowledge you get to apply.  Is crossing the street a matter of skill or luck?  Well, am I allowed to choose to not to cross the street if there are cars speeding by?  So, let's agree that you get to apply some basic principles in the various games.

One of the things we know about daily fantasy sports games is that you want to select players that are actually going to play.  It doesn't help you to choose a player on the Yankees if the Yankees aren't playing that day.  Thankfully, these sites limit the pool to teams that are playing.  It also doesn't help to choose a pitcher that started a game the day before, nor that he's been only a relief pitcher.  Indeed, when it comes to pitchers, your pool is restricted to the announced starting pitchers.

So, guess what DraftKings did in order to test the skill v luck:

Well, you can guess how that happened.  The simulator randomly chose a relief pitcher, and he'll randomly choose a bench player and so on.  In the NHL, it randomly chose the backup goalie.  In the NBA, since playing time is heavily skewed, it's critical to choose among the 5 or 6 players most likely to get minutes on each team.  If the pricing was fair, then we wouldn't have a problem.  That is, if Wade Davis was priced at 2000$, and if other relievers were priced at 1000$ or 500$, then that's fine.  But what we have is the site offering unfair prices, knowingly unfair prices (all relievers priced at no less than 4000$), on the idea that its participants are fully aware that you need to have players in your starting lineup.  And its participants are fully aware of that, and so, 99%+ of them will never choose a relief pitcher. Just like a blackjack player is fully aware that if you are holding 10 or less, then you must hit, since there's no cost to not hitting. 

If I saw this in a blog post, my response would simply be: dude... c'mon.

Sunday, November 15, 2015

Analyzing the DraftKings dollar values, part 4: when did they realize Yusmeiro Petit was an RP?

?As we've learned, DraftKings prices their players at around 550$ to 600$ per point, with a floor for pitchers at 4000$ (and nonpitchers at 2000$).  Wade Davis for example was always priced at 4000$.  Baseball Reference sortof has the scale right.  The 4000$ line on the right axis should match to about 7$ on the left axis.  From that standpoint, you can see that Wade Davis rarely is worth picking up.

Now, here's Petit, who in his career has split his time as SP and RP.  But in 2015, he had one start.  But he was priced as a SP for every single game, except for the last game of the season when he was priced at 4500$ (a reliever with a chance at being multi-innings).  Why did this happen?  I can only imagine that they don't have enough controls in place.  Now, it doesn't really matter, because if you have a player severely overpriced like Petit, then no one will pick him.  I'd like to know however how often he was picked.

Saturday, November 14, 2015

Analyzing the DraftKings dollar values, part 3: is it all about the pitchers?

?There were 11 pitchers who averaged at least 10,000$ in salary, and another 10 pitchers that averaged at least 9000$ in salary.

Among the 246 nonpitchers with over 100 games played (i.e., the guy in the starting lineup), six players averaged at least 5000$ and none topped 6000$.  Only 68 players averaged at least 4000$.  Basically, there is very little spread among nonpitchers.  The system is setup so that you live and die on whatever starting pitcher you select that day.

Analyzing the DraftKings dollar values, part 2: responding to pitcher talent level

?DraftKings seem to price their players at about 550-600$ per expected point.  If we focus on the top starting pitchers, this also holds.  Chris Sale for example had his usual great year, and it cost you 550$ per point to get him.  If you look at his price chart over the year, it's fairly flat.  He started where he ended up.  He went down as our estimate of his talent dipped following a few poor performances, and higher following a string of good performances.  A good way to see a forecast.

But the Cy Young favorites, Arrieta and Keuchel are not like that.  They cost around 450$ per point (after the fact).  At some point though during the season, it should have been apparent we had actual star pitchers here.  Did that happen?  With Keuchel, it happened on his 6/30 start.  From that point, to the end of the season, it cost you 550$ per point to get him.  On the other end, Johnny Cueto would have costed you 663$ per point if you kept picking him.  And DraftKings never responded with a lower price in light of new information. 

Analyzing the DraftKings dollar values, part 1

?Hans was kind enough to send me the DraftKings data they post over at Baseball-Reference.  Now before I do a more comprehensive analysis, I need to understand the data, so any expertise from the players out there would be helpful.

The first thing that jumps out is that if the player is on the roster, they get a dollar value.  So, Kershaw will cost you 10,000$ to 15,000$ any day you pick him... even if he pitched a complete game the day before. Therefore, you need to understand a smidge about baseball: starting pitchers are on rotation.  When I look at the data in part 2, I'm going to need to separate out pitchers from non-pitchers.

The second thing that jumps out is that 4000$ is a common price, followed closely by 2000$ as the next most common price.  Over 25% of the dollar values are set at 4000$.  The 2000$ figure occurs more often in September than in other months.  I'll look, but I'm certain those are for the September callups,

The third thing that jumps out, relative to the second thing, is the number of points earned by salary for that day.  LIMITED TO PLAYERS WHO PLAYED (which means you'd have to know who played), the points per game at each salary goes up the higher the salary.  This shows that, broadly, DraftKings prices well enough.... except at 4000$.  To give you an idea, here's the data close to 4000$:

•  6.6     3700
•  6.7     3800
•  7.0     3900
•  2.9     4000 
•  6.9     4100
•  7.0     4200
•  7.3     4300

So, the players who cost 3700$ earn you 6.6 points per game played (again, you HAVE to know he would have played).  The 3800$ players earn you 6.7 points per game.  As you can see, the 4000$ players are HORRIBLY mispriced.  So, unless you really really really know what you are doing, ignore all 4000$ players. 

This is a point I made about the 3000$ players for the NHL in a recent post, as I theorized that any kind of NHL player was priced at 3000$, the minimum level.  And it seems that the 4000$ players in MLB are like that.

Anyway, I'll be back after I hear more from you.

Friday, November 13, 2015

What is a game of skill and what is a game of chance?

?Ok, so I went to the Fanduel site to see how the thing works.  Before we can have this discussion, we need to understand what counts as skill and what counts as chance.

Back in the 1990s, Quebec launched Mise-O-Jeu, which is french for faceoff.  Quebec is HUGE on lotteries.  It was straightforward: you select three games, each of which has a payoff odds.  You multiply those odds.  If all three teams win, you get that payoff.  As long as the payoffs matches the skill level of the teams, there's nothing to arbitrage.  That's the key: arbitrage.  Well, that and the vig.  We'll talk about the vig too.

Mise-O-Jeu was so popular, it expanded to include European soccer.  That was an early problem, because they actually were taking bets in some cases on completed games!  I don't follow the betting scene, but I presume that Mise-O-Jeu is crazy popular, and I presume the odds are all reasonable enough that a non-hockey fan will win as often as a die-hard hockey fan.  That I think is the key, and that's what distinguishes it from being skill and chance.  Mise-O-Jeu is a game of chance, so under the purview of the government.

We can also talk about the vig.  Even if it is a game of skill, if you make the vig high enough (as all lotteries do, and Mise-O-Jeu is handled by the lottery commission in Quebec), and a skilled player will, at best, break-even.  But in any case, in this case, no skill, so, this issue isn't relevant here.

Let's go to Fanduel.  I went on the site, and selected NHL to see how it goes.  What they do is basically put dollar amounts on each player.  Fantasy dollars, and you get a cap for that day.  It kind of corresponds to the actual fantasy points, but it's really based on the FORECASTED fantasy points.  It's the right way to do it, because that's how everyone operates.  If you think you can forecast better than the next guy, there's an opportunity to show off your forecasting skills.  Marcel laughs at you.  So, whatever forecasting skills you think you have, it's going to be extremely minor.  You have to select 9 players, 2 at each of the three forward positions, 2 D, and 1 goalie. Fan Duel also limits your selections to players who are on those teams that day.  They make it very very easy for you.

Can a non-hockey fan do as well as a hockey fan?  No.  There are three reasons.

1. Goalies.  Just like in baseball, the key is the starting pitcher that day, so too is the key the goalie.  In hockey, one goalie will play virtually all 60 minutes and his backup will stay on the bench 90%+ of the time.  So, you don't want to pick the goalie who is not going to play that game. So, definite skill, or at least awareness of news.
2. Injuries or healthy-scratches.  Similar to above, but less serious, is that you want your player to play.  It's not as bad as with goalies because goalies get big time points just for being on the ice, while players who are likely to be healthy-scratches wouldn't have helped you either way (whether they play or not, they wouldn't have helped you anyway).  In addition, they give you little indicators of their health status, not to mention their games played so far. 
3. The minimum salary.  Ah, that's another source of skill.  No matter how bad you are, every player costs at least 3000$.  So, in terms of skill, you should NEVER select such a player (unless you are a hockey fan and have a really good reason to think this player is not a replacement-level fantasy player).  Because these 3000$ players could very well be worth 3500$ or 3000$ or 1000$ or negative dollars.  We don't know, because they put in a 3000$ floor.

So, how much skill is required in order to turn "chance" to "skill"?  Take Rock-Paper-Scissors, which you would think is a game of chance.  But, if you apply Game Theory, you should do better than 50/50.  If you presume your opponent is human, who is looking for patterns, you can beat that player by ignoring what he is doing and simply randomizing your throws (but make sure your opponent isn't told you are playing randomly).  The worst you can do is be 50/50.  The best you can do?  I don't know... maybe you'll win 55% of the time?  Is that enough to turn chance to skill?  I don't know.

But since almost nothing is pure chance, this will hinge on how often a "skilled" player can beat a random player.  And if he can beat the vig too.  See, if the vig is high enough that you need to win 55% of the time to breakeven, then a skilled player will not win money!

So, there you go.  There's definitely ways you can increase skill in Fanduel, and still keep the game going at the same high energy level.  Right now? The AG of NY can't really make the case that there's enough potential that a skilled player can lose to a random player sufficiently enough that this is a game of chance, not skill. 

Saturday, January 10, 2015

Quick way to rank starting pitchers for Fantasy Baseball

?Bobby had a good idea here.  Here are my comments:

1. I do the EXACT same thing with the z-scores as you did, including the part about ERA and WHIP.

2. You MUST include all pitchers, not just SP. More specifically, you have to include all pitchers that are presumed to be in whatever league you have. It doesn’t have to be perfect, just reasonable. This will have some affect on the z-score.

3. Why not simply put K, BB, and HR into a regression and let the regression tell you the weights? Especially since one of the 4 categories is itself strikeouts, I expect K to be overweighted compared to BB. But then BB is (a part of) WHIP, so, now it goes (somewhat) the other way! And who knows what kind of effect HR has on the whole thing. Maybe you get something FIPpy, and maybe not.

So, I'd love to see someone out there implement this.  Even something as simple as running a regression of forecasted K, BB, HR (of pitchers forecasted with at least 120 IP) against forecasted/published Fantasy dollars.  Someone want to do this for us as if today was Mar 15, 2014?

Sunday, April 13, 2014

Dirk Hayhurst, the new Curt Schilling?

?Yup, he's going for it.

Tuesday, February 11, 2014

Mock draft

?I must be really old.  I don't know what a "mock draft" is.  What is the objective of a mock draft?  Is it to try to crowdsource, but in a fun way? 

If so, the obvious control is an auction, with all bids must in increments of $1, and no sale valid unless a second bid is placed.  So, you can't start the bidding for Doc Halladay at 40$, and have no one else bid on him.

Related link.

Thursday, September 05, 2013

OOTP14 Review

For those who like simulation games, John gives it a review (as do the commenters after the article).?

Friday, July 26, 2013

WAR on Topps baseball cards

Well, isn't this cool news?  Doesn't specify if it's from Fangraphs or BR.com. 

Time to lobby for all the other metrics!  ?

Sunday, May 26, 2013

Table-top Baseball game: When the Grass Was Real

?I just came across this.  One of the game designers is good buddy Ted Turocy.  It's pretty impressive in its breadth and depth.  And, it does things just a little bit differently. 

Friday, May 10, 2013

Sim game OOTP 14 Review

Crashburn reviews the game.  I've never played one of these kinds of games.  And I haven't played any kind of sim baseball game for about 25 years. 

***

My preferred game was Jack Kavanaugh's Extra Innings, a well-designed intelligent, but simple, game.  I tried Replay Baseball?, and it was ok.  I tried Strat-O-Matic, and I thought it was good.  I don't think I've tried any others. 

My favorite of the "basic" game (no "real" players) was Vince Lombardi board game.  It was tremendous, because everything you needed was on one tabletop board.  Just a brilliant construction. 

My favorite of all board games in any sport was Canadian Pro Sports Action Football, which was brilliantly-designed, yet simple.  Its playability was tremendous.

None of the hockey games I tried (Canadian Pro, Strat) was any good for me.

Thursday, March 07, 2013

Fan-tasy rules idea that includes fielding: games played at premium positions

I've had this simple idea to create a category or points that is simply games played at premium positions.  That is, give two points for each game at catcher or SS, 1 point for each game at 2B, 3B, CF.  Make that its own category in a 5x5 league.  Or, in a points league, you'd have to set the scale so that you get reasonable results comparing CF to corner outfielders.

Anybody ever try this??

Saturday, February 23, 2013

The worth of SB, HR, and all other categories, in Fantasy Baseball

?An oldie, but always useful at this time of year. (post 60)

Friday, February 08, 2013

Strat-O-Matic

It's always fun reading stories about the game.

The trip to Glen Head is no longer necessary, even for the hardcore gamers. Strat-o-Matic's computer game (not an EA Sports-style video game, but a digital version of the board game, cards and all, with some basic graphics) can be quickly downloaded, as can a book of player ratings that gives gamers every scrap of information they need to hold a draft. "The lines are not what they used to be, but the interest is as high as ever," Richman said.

Bad weather is also likely to keep some gamers away on Friday, though nothing stops opening day. Thomas flew into a snowstorm just three years ago; Strat-o-Matic employees kept in touch with him by phone, noting the conditions would not only slow his travel, but could delay the shipment of the cards themselves from the printer. "When I arrived, I saw the truck [from the printer] parked," Thomas said. A storm in the early 1990s also jeopardized opening day, but did not cancel it. "Even if you lived in New York, it was a hassle getting here. But 75 people still got here," Richman said. "It was shocking."

Thomas began playing Strat-o-Matic in 1973 but did not start flying to Glen Head until 2005. "I thought it would be something to do one time, to have the experience of traveling to Mecca," he said. "It evolved into this monster where my league is almost depending on me." Thomas' eight sets will be distributed among the members of his league, a few of whom have continuously played together for nearly 40 years. Thomas and his friends will draft in March and conduct a 135-game season that will run through October, players meeting weekly at each other's homes.

Bender has been making his shorter trip for 25 years. In the past, he strove to be first in line, a quest that led to some ridiculous extremes. One year, he and three friends arrived four hours before the company doors opened, on a 15-degree morning with 15 mile-per-hour winds. "It was brutal. We were like four stray dogs," he said. Strat-o-Matic employees took pity on them, let them in the building, and gave them their card sets on the QT.

?

Thursday, February 07, 2013

Steamer for the Win, and Marcel continues to Place

RotoValue continues with testing forecasting systems.  Steamer continues to look mighty good, while Marcel continue to provides the minimum baseline, but that too often does well. 

In terms of testing, I think Geoff can improve the selection criteria.  He tried two different methods: one is all players forecasted, and another is only the common players.

The problem with the "all players forecasted" is that it helps those who forecast thousands of players, since most of those will have a value of zero dollars, and their 2012 performance (or non-performance) will value at zero dollars, so that helps on the average error.

Geoff figured that out, so he went the other way, and only looked at common players.  But, if someone forecasts Bryce Harper at 15 dollars, and someone else doesn't forecast him at all, he's dropped off, thereby preventing the guy with the great forecast from getting his due.

So, how do you fix this?  Well, if you have a league that drafts 250 players, then you go through each forecasting system, one at a time, and pick out the top 250 players, separately in each league.  Those who forecasted Harper get to count him at whatever value they counted him at (say 15 dollars).  If he was worth 18 dollars, then that's a 3 dollar error.  For someone who did not value Harper, well guess what, that system gets an 18 dollar error on him.