| RotoGuru Baseball Forum View the Forum Registry |
| |
| 0 |
Subject: Investigations in SW Price Changes Posted by: Richard - [1719313] Tue, Apr 17, 09:56 I'm sure that, like me, most of you were surprised by the small drop that Pedro Martinez had on the April 15th price change. On the day after his start and on a day lots of manager's should have traded out of him he only dropped -190K instead of the expected drop of around -500K. This surprize motivated me to do a bit of investigating. Luckily I was able to look at trade patterns just before SW hid that information behind their increased security measures. For those of you who need a background refresher on SW's pricing model, please see Guru's initial explanation Peeking Behind the Pricing Curtain, May 24 1999 . Somethings have changes since then, like price gravity, but the basics remain the same. The formula is linear and based on the concentration of trades (percentage of trades made) and not on the total volume of trades made. For the past two or three years, I've looked at SW pricing and how it relates to trades. The simple linear relationship of price changes to concentration of trades has been a remarkable predictor of price changes in SW Baseball, SW Football, SW Hockey and even SW Golf. It's no different this year except for a few price anomolies centered around high trade concentrations for Pedro Martinez. First some brief data and discussion. This year's baseball re-pricing can be explained by a simple linear relationship. If you divide the net trades (buys - sells) for an individual player by the total trades made each day (buys + sells) and multiple that percentage (concentration of trades) by $30K you get the price change. this relationship holds for days with high trade volumes and days with low trade volumes. I have captured the trade activity for 538 random teams and have based my investigation on this small sample of teams making the assumption that these random manager's have a trading pattern similar to the entire universe of SW managers. For example On April 15th these manager's made 277 pitching trades. Wade Miller was bought by 11 managers so 3.97% of the trades (buys + sells) were concentrated on him. My pricing model predicted an increase of $119K and he went up $120K. April 11th was a very heavy trading day with 522 pitching trades made. Mike Mussina had a net trade total of -37 or -7.09% which translates to a predicted drop of -$213K. In reality he dropped -$220K. April 14th was a very light trading day with only 160 pitching trades made. Wade Miller was a popular pickup capturing 9 net buys or 5.63% of the total trades. His increase was predicted at +$169K and he had a +$180K actual price increase. These examples represent a few selected data points to illustrate that price changes are relative to the concentration of trades and not to the over all volume of trades. The relationship of a $30K price increase or decrease for every 1% concentration of trades holds pretty steady regardless of total trade volume. So what happened on April 15th with Pedro Martinez. He had a remarkable sell-off with 54 of my small sample of manager's selling him for a concentration of -19.5% (-54/277). This should have equated to a drop of -$585K. So what did SW do? It looks like either of two things happened. First, they might not have captured all of Pedro's sells (his drop of -$190K suggests that only 6.33% of the total trades were Martinez sells). Secondly, they might impose a price movement cap when trade concentration gets real high. On most days I see trade concentations of less than 10% but on April 15th Pedro had a remarkable concentration of -19.5%. Maybe SW has instituted a price movement cap when trades are highly concentrated. BTW - if you are still curious about this, on April 9th Pedro had another big drop in price (-$470K). His trade concentration on that day was -12% and I was predicting a drop of -$360K. Richard | ||||||
| 39 | slimer ID: 9150216 Tue, Apr 17, 15:34 | i just clicked on another team in one of my divisions and i see smallworld has went back to their old ways....u can see all the info for other teams now (rv, trades remaining and what not) i figured they would do it eventually | |||||
| 40 | KrazyKoalaBears ID: 431156218 Tue, Apr 17, 15:44 |
KevinL, no arguments here about running everyone through the same formula, but it just doesn't seem to be what SW is doing. As Madman pointed out, unless they fixed something, Helling's first day drop gives a tiny bit of evidence that large gains/losses are possible. If it happens again, then we'll know for sure (sorta). At this point, I trust nothing about SW and am going to be looking at other pitching rotations and my hitters for money. I'm seriously considering holding Pedro and just forgetting about his spot in my roster given the obvious targeting of Randro, whether by a single formula or multiple formulas. I'm going to have to give it more thought, but I think I'm heavily leaning that way. | |||||
| 41 | jumpball ID: 33050298 Tue, Apr 17, 15:47 |
Let me mention one more thing which lends support to Wammie's theory. The only large price drop for Pedro occurred on the day when both he AND Randy pitched. That means that there would have been no Randy buys that day. And thus no dampening of the price change. Richard -- excellent work, as usual. Thanks for sharing. | |||||
| 42 | Sanchez ID: 11311921 Tue, Apr 17, 16:14 | Good observation jumpball. | |||||
| 43 | The Left Wings ID: 2131321 Tue, Apr 17, 18:03 |
Jumpball, are you saying that it's Randro that's targetted? Since Both of them pitched that day, there was no Randro movement and there for Pedro's drop was not dampened? Man that'd be stupid of SW to do so... So much about free-trade, and non-intervention from the "government". | |||||
| 44 | Wammie ID: 437541618 Tue, Apr 17, 18:28 | LW, that is exactly what I am saying. On that day, all of the Randy and Pedro changes were to other pitchers, not each other. on that day, the Randro was not attacked, because it didn't exist. I agree that it is a crock for SW to do that, if that is indeed what they are doing. | |||||
| 45 | Strike One ID: 39252299 Tue, Apr 17, 19:05 |
great stuff richard. i don't really care how smallworld arrives at the price changes, as long as they are consistant. i'm not mad at smallworld for dampering the price changes for pedro and randy, but to have pedro's price drop 490K one day, then 6 days later have it drop on 190K, when practically the same amount of sells were made is just wrong. they can not keep changing the reprice algorithim, well i guess they can, but it makes the game unfair and outright fustrating when the rules change. I'm not as worried about randy/pedro price changes as all the other players. it would make sense that if there were a total of 1,000,000 SWD worth of gains that there would be around 1,000,000 SWD worth of losses. but this has not been the case in the last week or so. when you sell a shortstop you must buy a shortstop, you can't by another catcher or outfield. if you want a new outfielder you have to sell an outfielder. so the net change in price changes for each posistion should be close to zero. i'm not sure where i'm going with this but whatever happens, smallworld needs to stick with a repricing plan that works and makes sense and not change it whenever they feel fit. | |||||
| 46 | The Left Wings ID: 2131321 Tue, Apr 17, 19:41 |
Dude Strike One, you should really read the whole thing first. There are a couple of things that we've discussed that you totally missed. #1. We don't think they kept switching algorithms. We think they specifically targetted the "sell Pedro, buy Randy" and vice versa trades. Other "sell Pedro, buy other guy" trades were not affected. #2. You forgot about the buys from all the new teams who only have buys and those teams (for whatever strange reasons) chose to go invalid. | |||||
| 47 | Madman ID: 29246911 Tue, Apr 17, 22:20 |
?? Left Wings -- It's clear that they are indeed switching algorithms. On April 9th, Pedro dropped by 125% of what he should have. Yet, 6 days later, Pedro dropped by only 33% of what he should have. The only way that this is NOT a change in the algorithm would be if you believe: a) the linear relationship found by Richard holds for small trading amounts, b) between 10 and 14%, all trading is exaggerated, c) above 15%, all trading is dampened. That seems utterly non-sensical to me. | |||||
| 48 | Strike One ID: 39252299 Tue, Apr 17, 22:28 |
LW, after re reading my post i think that i covered point #1, and i did ignore point #2 but i don't think that that has a very big effect in the significance of the price changes. btw, i did read the whole thread, and it has amazingly great posts by everybody, and this is definetly one of the best threads of the year. | |||||
| 49 | The Left Wings ID: 55105317 Tue, Apr 17, 23:28 |
I'm leaning towards believing that the 2 or 3 guys left in SW are not smart enough to find a non-linear way to calculate price change. I think they'll probably be hard-coding the Randro trades. We all agree that they're idiots in SW, right? Well, maybe except Don Mathis, who seems pretty normal to me. | |||||
| 50 | KevinL ID: 10417811 Wed, Apr 18, 09:26 |
Richard, can you give us the numbers, for the Pedro dates in question, of Randy buys? For example, 36 Randy buys on 4/15 would put Pedro on the right spot, IF Randy buys are subracted from Pedro sells. | |||||
| 51 | KevinL ID: 10417811 Wed, Apr 18, 09:45 |
Madman, not so fast. April 9 was the day Randy & Pedro both pitched, right? So if Randy buys were subtracted from Pedro sells (as suggested here), it's also possible that randy sells were added (opposite of subtracted) to Pedro sells, right? To modify KKB's formula. PS = Pedro sells PB = Pedro buys PT = Pedro Today PY = Pedro Yesterday RS = Randy sells RB = Randy buys RT = Randy Today RY = Randy Yesterday KKB suggests that: ((PS-RB)/TRD) * $30k Well, if you assume RB = RY - RT, and substitute (RY - RT) where you have RB, you could possibly have a negative buy (a sell) which would increase (not decrease) Pedro's price drop. Possible? | |||||
| 52 | Madman ID: 29246911 Wed, Apr 18, 09:59 |
Pedro and RJ both pitched on April 8. Plus, as you recall, RJ got shelled on April 8, although he was going to pitch before Pedro later. Why in the world would they ADD Pedro and RJ sells and apply them to Pedro, but not do this for RJ? There are an infinite number of formulas that actually do match the data as given. But I would suggest that all of them are so improbable as to not warrant consideration. SW goofed. I follow the KISS principle. | |||||
| 53 | KevinL ID: 10417811 Wed, Apr 18, 10:12 |
On 4/9, the day after both pitched, Pedro dropped 470, Randy dropped 160. We know Pedro was sold on 12%, but his price change corresponds to 16%. Is it possible Randy was bought on 4% of the rosters? (call it blind Randro if you like) Meanwhile, to get his 160k drop, Randy sells should have been just over 9 %, if we assume 4% buys. We know RJ got shelled, and that his next start was in Coors Canaveral, but we don't move the market, the masses do. | |||||
| 54 | Richard ID: 1719313 Wed, Apr 18, 10:15 |
KevinL - I think you might be pushing the numbers a bit hard since my trade sample is quite small (only 538 teams) and thus these numbers can only be an approximation of the true trading patterns, but since you asked, here's what I think you want: April 15th I found a total of 137 buys and 140 sells for 277 total trades (as I count them). Randy had 5 buys and 3 sells for a net of +2. Pedro had 0 buys and 54 sells for a net of -54. April 9th I found a total of 181 buys and 180 sells for 361 total trades. Randy had 5 buys and 20 sells for a net of -15. Pdero had 3 buys and 44 sells for a net of -41. April 2nd I found a total of 152 buys and 154 sells for 306 total trades. Randy had 70 buys and 0 sells for a net of +70. Pedro had 0 buys and 80 sells for a net of -80. Enjoy the number and let me know if you need some more. Madman you asked earlier about my correlation and it's goodness of fit. I have a very small subset of price change and trade concentration data. I have taken 90 different price change points from the Top Gainers and Losers lists for April 9th thru April 13th. I then asked Excel to fit a trend line thru the data forcing the line to go thru 0. The resulting equation looks like: price change = 31.486 * trade concentration with a r-squared of 0.92 The correlation is quite high but most of my points are out near the extremes of both positive and negative price changes. The data doesn't include the dampened price changes that from either April 2nd or April 15th. I just tried to use a few days of "normal" price changes to develop my correlation. Remember, I'm a geologist by training not a statistician so don't join up with Sludge and tag-team me on my stats;-) Richard | |||||
| 55 | deepsnapper ID: 36053257 Wed, Apr 18, 10:43 |
Thanks a lot Richard. Your usual excellent detective work. I see a sampling for table_id=19 in your future. ;-) This topic has generated lots of great dialogue including some which is giving me a headache trying to follow. Richard, I think you and Madman are dead solid perfect on your suspicions. They're manipulating the market when the % of trades and volumes hit target numbers they've selected as needing to be dampened. Until we have another "Ramon" incident, the "dampening effect" probably won't effect anyone other than Pedro or Randy because no other combination of trades will meet the criteria. Maybe a RanBro, MadBro or MadDro will trigger the "govenor", but I doubt it. I believe SW is targeting high trade concentrations and we're only seeing it's effect on RanDro because so many managers are "doing the randro shuffle" because of it's success in the past. On a scary note, what if their intentions are to make hitter trades more valuable than pitcher trades by "tricking" the system in favor of HT? Did anyone see the Velarde gains this week? wow! I see a Guru Jr. in the near future for Richard. ;-) | |||||
| 58 | Tranceformer ID: 26224814 Wed, Apr 18, 11:09 |
a .92 correlation is actually not great. its R^2....so the actually correlation percentage is actually the square root of .92..making it 95 percent....95 percentage is right on the cusp of being significant in statistic terms(2standard deviations within the mean)....in statistics a R^2 of around .98 or .99 is what you're looking for in order to feel that your model fits the system well....feel free to drop a huge outlier or two in order to get a better fit.... if many of your points lie either on the negative or positive side, you might want to try fitting each set separately....(one fit for the gains, one fit for the losses)...this would fit into the "sells are worth more than buys" theory that many SW managers hold...... | |||||
| 65 | biliruben ID: 231045110 Wed, Apr 18, 12:19 |
Transformer - Just to warn you, sludge is a statistics professor. He also happens to be right, which is not a given among professors. ;) He may have calculated it by hand once or twice. r-squared is he proportion of the predictable variance. The predictable variance divided by the total variance. It is not the same thing as p. | |||||
| 67 | Strike One ID: 39252299 Wed, Apr 18, 12:25 | i'm going to take AP stats next year in school and i'm already reading looking forward to it. look at the stuff i could figure out with that knoledge! :) | |||||
| 79 | Sludge ID: 1440310 Wed, Apr 18, 13:39 |
I have withdrawn all of my replies to Tranceformer. I'd be more than happy to argue all day long about R^2, but I don't desire to be party (any longer) to reducing this thread to the level it's headed right now. | |||||
| 81 | Madman ID: 29246911 Wed, Apr 18, 14:03 |
Geez. I came back to this thread expecting some more good stuff, and, oh man . . . we need some help from the mods, I think. At any rate, Richard 54, thanks for the clarification. I was just asking to get another idea of what you had already stated anecdotally (that your predictions were indeed very close). Actually, given the measurement error inherent in your methodology, that's darned good, I think. Without cranking the numbers, I feel comfortable that the anomalies you've identified are indeed anamolous. As the season progresses, I think we'll get a better feel for whether or not they are targetting Randy and Pedro specifically, or all high-percentage movers (my guess is the second -- and that they screwed up on 4/2 -- but who knows??). | |||||
| 89 | Richard ID: 1719313 Wed, Apr 18, 15:11 |
Madman - this thread did take an unfortunate turn after post #55. A sad state of affairs, definately a sad state of affairs. Alas, I hope we all learn from this, but I have my doubts about some of our younger members. It makes me kind of glad I was never a National Merit Scholar;-) Richard | |||||
| 98 | deepsnapper ID: 32319720 Wed, Apr 18, 17:16 |
PD - I had the same thing in mind. As long as it doesn't have a table, I can clean it up. I need work with tables. Digging around looking at the price changes in Hardball, I saw where SW added a 2 week price history to the player's page like we had a few years ago. It made me go check the free game. Here's the Randy Velarde price history page. Back when I had to implement program changes, we had a thing called "Change Management" we had to follow. It helped out a lot with customer service. D'Oh! | |||||
| 99 | Madman ID: 29246911 Wed, Apr 18, 18:30 |
Continuing R^2 discussion Mods, you can feel free to delete this post if you wish. I wanted to self-edit my previous post in which I gave this link, but figured I should leave the link. . . | |||||
| 100 | slimer ID: 2644430 Wed, Apr 18, 23:31 |
we will have a better idea (at least i hope) come friday morning....everyone should have trades to get out of him this time i still think i will just keep him this time...i got bit his first big drop then made sure to sell him after his last start and we all know what happened... i have a feeling i am going to hold him and everyone will sell him this time because they have trades to sell and i will get a big chunk taken out of my ass again....oh well | |||||
| 101 | Slow Stick ID: 133391823 Wed, Apr 18, 23:39 |
slimer please sell! I was going to hold and I cannot take the hit you cause when you do not sell! lol Slow Stick | |||||
| 102 | swank ID: 36437422 Thu, Apr 19, 00:22 |
Dude.....my brain hurts now!!!!! I need a beer and then I'm trading Randy for Pedro. Swank | |||||
| 103 | deepsnapper ID: 36053257 Thu, Apr 19, 06:43 |
Depending on the accuracy of Rotowire's PP, Pedro is scheduled to pitch next Wednesday vs Minn after today. I'll either move to Mad Dog for his Sat start vs Philly or RJ for his Mon. start vs Fla. I'm not holding Pedro for a week. Money considerations or not, I'd rather have the extra start this week. | |||||
| 104 | Pond Scum ID: 54420321 Fri, Apr 20, 11:09 |
-$520,000 WHAT dampening effect? I was tempted to go to a hold Pedro strategy, glad I didn't. | |||||
| 105 | The Left Wings ID: 2131321 Fri, Apr 20, 11:10 | Looks like they damp Pedro every other start... | |||||
| 106 | jumpball ID: 33050298 Fri, Apr 20, 11:32 |
Looking closely at toady's price changes indicates that was no change in Randy's price. Therefore there was no way to "dampen" the Randro effect. Next week there are two days between Randy and Pedro (instead of 1). If there is dampening going on, we'll see smaller numbers then. | |||||
| 107 | GoatLocker ID: 26058212 Fri, Apr 20, 11:38 |
I'm not sure what is going on, but if you look at everything, it appears everybody went from Pedro to one of three options - Maddux / Schilling / KB. Then the next move will be from whomever to RJ. My guess is that we will see a big RJ positive. Sure does look like hard coding of Pedro > RJ RJ > Pedro. Anyway, that is my $.02 Guess the best option is to wait and see, and I'm sure glad that I didn't hold Pedro like I was tempted to do. Cliff | |||||
| 108 | Roo ID: 5405365 Fri, Apr 20, 11:51 |
I'm sure there are plenty of others, like me, who will be moving from KB (or Schilling) into Maddux tomorrow, before everyone moves to RJ. Hold on tight! | |||||
| 109 | Wammie ID: 20039259 Fri, Apr 20, 11:58 |
I am going to Maddux tomorrow. I was goign to buy and hold, but with him making this much money today, and then he will turn around and have a good day tomorrow. Maddux is getting set up for a big fall after his start. I am going to continue to hold RJ after his start this week as well. any one who owns him, will likely go to pedro. should have a "dampened" effect. | |||||
| 110 | ChicagoTRS ID: 4324316 Fri, Apr 20, 12:31 |
Funny that RJohnson did not move in price at all today. The ten people in my division...6 moved from Pedro to Randy Johnson. Obviously a small sample size but I would bet there was a significant amount of people who did do this. Seems like Randys price was dampened. Could it be that SW just targets the trades between Pedro and RJ / RJ and Pedro? Probably many more people went Pedro to ??? and then will go to Johnson so maybe that is why we saw a significant drop in Pedro because many of his trades were to Maddux/Schilling/Brown and not Johnson? | |||||
| 111 | ChicagoTRS ID: 4324316 Fri, Apr 20, 12:49 |
I would be interested in a sample of yesterdays trades? Especially on the pitching front. Obviously everyone dropped Pedro...but I wonder how many went to each pitcher. Just seems hard for me to believe that RJ had a zero gain...I know some significant percentage must have went "randro". I am starting to think SW is only targeting the trades between these two pitchers. | |||||
| 112 | Baldwin ID: 23311413 Fri, Apr 20, 14:15 | Looks to me like the new rule is...never hold Randy or Pedro if there is a big gap between the next Unit/Pedro start. Holding them when there is only a day or two till the next Unit/Pedro start will no longer kill you. | |||||
| 113 | Baldwin ID: 23311413 Fri, Apr 20, 14:17 |
Let me state that more precisely. Never hold Randy or Pedro if there is a big gap until the next Unit/Pedro start. | |||||
| 114 | Baldwin ID: 23311413 Fri, Apr 20, 14:21 | This is based on what I assume to be the new state of affairs. That only trades directly from one to the other are price change muted. If the masses are more likely to do some other tradeout then expect a huge drop in price. | |||||
| 115 | Baldwin ID: 23311413 Fri, Apr 20, 14:24 | Trade efficiency be damned, I am falling more and more in love with Unit/Pedro/best alternate stud with the best alternate stuck in the long gap. | |||||
| 116 | The Left Wings ID: 2131321 Fri, Apr 20, 14:30 |
Yup. I'm going for the three-man leaf or a 3-4 rotation. | |||||
| 117 | Cheeseheads ID: 533401811 Fri, Apr 20, 14:31 | I actually bought RJ and held him. I skipped Pedro this week all together since I already had Maddux just to see what the price of RJ would do. Looks to me like holding RJ may not be a bad choice for awhile and just trade Pedro-Maddux-Brown or whoever else is pitching. Looks like the money is to be made by rotating RJ or Pedro to the other pitchers not just purely RANDRO. | |||||
| 118 | Richard ID: 1719313 Fri, Apr 20, 14:53 |
ChicagoTRS - in my look at roster changes between April 19th and April 20th, I saw these net ownership changes Pedro Martinez decreased -67 from 242 to 175 and Randy Johnson increased 2 from 37 to 39 Greg Maddux increased 30 from 40 to 70 Curt Schilling increased 21 from 100 to 121 Kevin Brown increased 17 from 23 to 40 Brad Radke increased 8 from 33 to 41 I saw a total of 280 "trades" (actually I summed the absolute values of the net changes for each pitcher to approximate total buys and sells) and today's (April 20) price change was remarkably linear price change = $22.83 * (net trades/total trades) with a R^2 of 0.9705 Please note that I had to treat total trades differently than I did in the previous analysis since I can now only work with changes in total ownership for each pitcher and not in individual buys and sells anymore. That is why the slope of my linear trend line is different than the $30 I stated above. Richard | |||||
| 119 | Sludge ID: 1440310 Fri, Apr 20, 14:55 | Oh great... Not another R^2, Richard? :) | |||||
| 120 | ChicagoTRS ID: 4324316 Fri, Apr 20, 15:05 | Richard...thanks Using those numbers todays price movers make sense. I guess the real test comes next week when many people should be moving from Randy to Pedro. | |||||
| 121 | Cheeseheads ID: 533401811 Fri, Apr 20, 15:09 | The question is that next week Pedro has 2 starts and will the masses hold him for both. I may just hold Pedro for 2 and rotate Maddux and Randy. The masses may be doing the 2 starts per week thing. | |||||
| 122 | Baldwin ID: 23311413 Sun, Apr 22, 08:52 | BUTT | |||||
| 123 | KrazyKoalaBears ID: 51521713 Sun, May 20, 10:08 | BUTT. Any other ideas? ;) | |||||
| 124 | Questor ID: 404362111 Mon, May 21, 17:47 | BUTT | |||||
| Rate this thread: | If you wish, you may rate this thread on scale of 1-5. Ratings should indicate how valuable or interesting you believe this thread would be to other users of this forum. A '5' means that this thread is a 'must read'. A '1' means that this is a complete waste of time. If you have previously rated this thread, rating it again will delete your previous rating. If you do not want to rate this thread, but want to see how others have rated it, then click the button without entering a rating, or else click here. | ||||||
| |||||||
Post a reply to this message: (But first, how about checking out this sponsor?) | |||||||