| RotoGuru Hockey Forum View the Forum Registry |
| |
| 0 |
Subject: SW Repricing Algorithm Problem Posted by: KrazyKoalaBears - [30671612] Tue, Oct 10, 21:27 There is an obvious problem with Smallworld's Hockey repricing algorithm. When the top 15 gainers in a day range from $150,000 to $40,000 and the top 15 losers only range from -$30,000 to -$20,000, it doesn't take a rocket scientist to see something is up. Anyhow, my point is not to try and use the top 15 gainers/losers to show there is a problem, rather to explain what piqued my interest to do what I did. So what did I do? Here goes... According to Smallworld, the repricing of a player is determined by the number of buys and sells for that player for a given time (Baseball: 1 day, Football: 1 week, Hockey: 1 day). Furthermore, buys are not as weighted as sells due to the buys occuring from new teams drafting their players after the first game, thus adding buys without adding sells. Lastly, Smallworld's repricing also has "gravity" that causes a players price to drop slowly in the event of inactivity (as they put it, like an elevator, though I've never been on an elevator that falls over time). So what does all this mean? If you take it at face value, it means that the sum of the changes in price for all players should be negative. Because buys and sells should equal out due to weighting, the only other factor is "gravity". Since "gravity" is negative, the sum should be negative. But just how much negative should the sum be? That's where I come in. First off, since you have to trade one player for another of the same position, all figures are positional based. I looked at the following price changes and made an interesting discovery: FS Baseball Pitchers and Batters (Last Day), MS Baseball Pitchers and Batters (Last Day), Football QB, RB, WR, TE, K, DEF (Today's Update), and Hockey C, W, D, and G (Today's Update). So what did I find? This is what I found:
Now if you look at the chart, you see some trends. First off, you can see a clear average amongst the respective games. Baseball's average change per player per pos is -$5,680. Football's average is -$17,720, while Hockey's is a meager -$326. Now you can assume that Smallworld increased the price change effects for Football since it's a weekly price change (or lowered the price change effects for Baseball/Hockey because of daily repricing) and that leaves a comparison of Baseball and Hockey. In Baseball, your average player would lose $5,680 per day ($39,760/week), while in Hockey your average player only loses $326, or $2282/week. You lose more per player per day in Baseball than you would lose in 2 WEEKS in Hockey!!! My point? There's something wrong with the repricing algorithm and it's plainly obvious. While this is seems to be ok for regular teams (makes it easier to gain money, or harder to lose money), it's just the opposite for LRV teams and makes them practically impossible and a guess at best. But also, as we have seen in the past week and today, it's not benefiting the Gainers like one would imagine. A high today of +$150,000 from Shane Willis? The day after receiving trades?!? Something's wrong and Smallworld needs to fix it. Are the miscounting Sells? Is the "gravity" not working? Is the algorithm just messed up? Is it the last laugh of a laid off employee? I don't know, but it needs to be investigated. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 1 | The Left Wings ID: 55901022 Tue, Oct 10, 22:04 | My explanation is that there are 400 other players who were dropped by gravity ($20,000) and didn't show up. This amounts to about $8M, which would cancel out those players who rose in price. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 2 | KrazyKoalaBears ID: 249571011 Wed, Oct 11, 08:58 |
What do you mean by "didn't show up"? The statistics I provided include ALL players listed for each sport. Furthermore, why would CuJo go UP $20,000 the day after posting a -5 SWP? Were that many people anxious to pick up a goalie who netted -5 SWP the night before and didn't play the night they picked him up? | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 3 | Urmas ID: 5782874 Wed, Oct 11, 10:41 |
Another explanation could be, that at the dates you used there were more people joining Hockey game and less joining Football of Baseball. So it may be the buys of new fantasy players that make the difference. Or maybe not... Somethings rotten in SmallWorld. Just my $0.02 | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 4 | KrazyKoalaBears ID: 249571011 Wed, Oct 11, 11:43 |
Urmas, but as I pointed out, SW is supposed to be taking this into account in their pricing algorithm. From SW: "First of all, a sell actually moves the elevator down more than a buy moves the elevator up. That's because every day, new owners join the game and draft new teams. Each new team has a bunch of buys and no sells. So, at the end of the day, our market ends up with more buys than sells. To adjust for that, we make each buy have less "pushing power" than each sell." Also, yesterday there were around 138,000 teams (From WWR) and now there are 141,000 teams. So let's give SW the benefit of the doubt and say 5,000 teams joined from freeze to freeze. That 5,000 teams would be 3.5% of the "pool". Now let's assume that 75% of those 5,000 teams draft a team (I'm being VERY generous here). That means that 45,000 players were "buys" from this group of teams alone. Now assume that each of the 136,000 teams (since I'm giving SW 5,000 new teams) use 1 trade the day that trades were issued. That makes 136,000 "sells" and 181,000 "buys". Since 45,000 trades were from drafting, this 45,000 can be termed "unexpected buys" (I will assume for this that SW doesn't take these into account). So the 181,000 "buys" are 133% of the 136,000 that was "expected". So, using this explanation, the sum of changes in price would be raised by 133%. Working backwards from this, you would have the following Avg price change per player per pos: This is still not "normal" according to the past. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Post a reply to this message: (But first, how about checking out this sponsor?) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||