Checking the TradeSports predictions--posted by Tony Garcia on 11/09/2006
I began writing a series of posts (sorry, they are of the lengthy variety) analyzing what happened, why, what it means and possible paths for the next 2 years. However, I wanted to interrupt that to post a quick report about the TradeSports predictions (which King thinks I should stop using).
Sadly, this is the last election cycle that TradeSports predictive nature can be used. The GOP in Congress, on its last night before being thrown out of power, passed some "we know what is better for you than you do" legislation known as the Online Gambling Ban. As a result online poker cannot be funded through American banks or credit cards. This effectively bans funding online futures trading such as TradeSports.
Back to the accuracy of the predictions.
Of the 33 Senate races TradeSports listed only 7 of them at less than 90% certainty. The seven races were MD, MO, MT, RI, TN, VT (don't understand that one) and VA. Those include 5 of the 6 races decided by less than 10 points (AZ being the odd one out).
TradeSports Senate Record: 33 CORRECT out of 33 RACES (100%)...and a not too shabby forecasting record on the closeness of the races. 6 of the 7 races (85.7%) TradeSports predicted as being close actually were. As far as the races that actually were close (I expanded that definition to being decided by 8% or less) TradeSports called all of them less than certain...that is 5 out of 5 (100.0%).
TradeSports Dem House Pickup Record: There were a series of 10 contracts that were giving an overall prediction of how many seats would turnover. The best way to evaluate TradeSports' record is to find where the 50% was. It fell between a turnover of 19 1/2 seats to 24 1/2 seats. Assuming the 10 undecided races hold the results being reported the actual pickup will be 29 (and that includes the Independent that went to the Democrats...as listed in the rules of the contract). So, TradeSports undershot the turnover by 5-10 seats. Not bad considering the increments of the contracts were 5 seats.
TradeSports Governor Races: I did not follow these at all, so researching the prices was a bit of a task. The results (just to make it a bit more difficult for TradeSports) are based on the predictions from November 1 through today. The results are also assuming that Pawlenty is confirmed the winner in MN. The record: 35 CORRECT out of 36 RACES (97.2%). The lone miss was MN. The market was very certain of these results as early as November 1.
* 30 of 36 (83.3%) were never dropped below 90%.
* 32 of 36 (88.8%) were never dropped below 80%.
* 34 of 36 (94.4%) were never dropped below 70%.
* 35 of 36 (97.2%) were never dropped below 60%. The lone one, of course was MN, which TradeSports predicted as a loss.
Find another pundit or poll that got 100% of the Senate races correct AND 97% of the Governor's races AND 86% of the "only these Senate races are less than certain" correct. Once you find that list see which ones also predicted the Democrats would pickup at least 24 seats.
Where the TradesSports market did go wrong was listing the GOP with a 72% chance of holding the Senate. W-R-O-N-G.
The other place were the parlays. When taken together the prices indicated overpricing by 11%. They were predicting the GOP to hold the Senate and the Dems to grab the House.
All in all, the predictive nature of TradeSports on a single event (each single race) basis is still incredibly high. Governor & Senate races combined was 98.6%.
Better than the pollsters!
(Too bad the GOP killed this tool for future use with their pro-nanny state legislation.)