ΠŸΠΎΠΌΠΎΡ‰ΡŒ Π² написании студСнчСских Ρ€Π°Π±ΠΎΡ‚
АнтистрСссовый сСрвис

Calibration of predictions in sports

ΠšΡƒΡ€ΡΠΎΠ²Π°Ρ ΠšΡƒΠΏΠΈΡ‚ΡŒ Π³ΠΎΡ‚ΠΎΠ²ΡƒΡŽ Π£Π·Π½Π°Ρ‚ΡŒ ΡΡ‚ΠΎΠΈΠΌΠΎΡΡ‚ΡŒΠΌΠΎΠ΅ΠΉ Ρ€Π°Π±ΠΎΡ‚Ρ‹

I n an efficient market, based on such prediction model should perform equally with simply using the difference in propagation predict endpoint. W e believe that this will be as follows: a 5-fold cross-validation, standard deviation 13.3 points for this model is identical to the RMSE spilled alone.5 In addition, despite the differences in the fcc structure and TradeSports Vegas markets, both… Π§ΠΈΡ‚Π°Ρ‚ΡŒ Π΅Ρ‰Ρ‘ >

Calibration of predictions in sports (Ρ€Π΅Ρ„Π΅Ρ€Π°Ρ‚, курсовая, Π΄ΠΈΠΏΠ»ΠΎΠΌ, ΠΊΠΎΠ½Ρ‚Ρ€ΠΎΠ»ΡŒΠ½Π°Ρ)

Π‘ΠΎΠ΄Π΅Ρ€ΠΆΠ°Π½ΠΈΠ΅

  • 1. Introduction
  • 1. The essence of forecasting
    • 1. 1. The concept of forecasting
    • 1. 2. Forecasting Methods
    • 1. 3. Factor as a predictor of sport achievement
  • 2. Description of the three methods of forecasting
    • 2. 1. Market Forecast
    • 2. 2. Tipsters
  • 3. Description of the Data
    • 3. 1. Data Set
    • 3. 2. Calculations of the Prediction Market Forecasts
    • 3. 3. Calculations of the Betting Market Forecasts
  • 4. Forecast Accuracy of Three Methods
    • 4. 1. Evaluation Criteria
    • 4. 2. Forecast Accuracy of Each Method
    • 4. 3. Forecast Accuracy of Combinations of the Methods
      • 4. 3. 1. Accuracy of Weighting-Based Combined Forecasts
      • 4. 3. 2. Accuracy of Rule-Based Combined Forecasts
  • 5. Results
    • 5. 1. Football
    • 5. 2. Baseball
  • Conclusions
  • References

Consistent with past empirical studies and theoretical arguments, the Vegas and the TradeSports markets are the best performers, both having an RMSE of 0.

46. At the other extreme the baseline model is the worst performer, with an RMSE of 0.

49. The performances of the remaining strategies lie in between that of the markets and the baseline model: Probability Sports, the win-loss model, and the filtered polls all have an RMSE of 0.

47. T o aid interpretation of these results, and also to ensure that the markets are not handicapped by our conversion of spread to probabilistic predictions, we consider the complementary problem of predicting the final point difference between the playing teams. F igure 2 shows that the standard deviation in this case is in the range of 13.3 to 14.5 for the markets for the basic model. O n average, that is, market forecasts differ on the difference between the actual point of about 13.3 points, and forecasts from the base model off 14.5 points.

I n general, the ordering of these prediction methods is not surprising: prediction markets beat patterns and polls, and all methods to beat the baseline. W hat is surprising, however, is that the various mechanisms so little different: the difference in the prediction at the end point, the gain loss model, which recall the parameter has only three-points is only 0.4 (3%) is worse than markets, sports and probability is only 0.1 points (1%) less than in the markets. F.

igure 3 shows the difference between the market and the Vegas model of win-loss from 1978 to 2008, the year of similarity in the performance of these prediction methods, moreover, not because of any obvious anomalies in the markets. T o check the obvious inefficiency of the market, we advance the difference in the predicted endpoint in each game by the model, which includes the market spread along with several other features. I n particular, we fit a linear regression model, where the spread of the market predicted a distribution point, Xspread> 0 is a dummy variable indicating the spread is greater than zero, and Hi and Ai are fictitious, indicating which teams play in this game. I.

n particular, the model corrects a systematic error that depends on which teams are playing. I n an efficient market, based on such prediction model should perform equally with simply using the difference in propagation predict endpoint. W e believe that this will be as follows: a 5-fold cross-validation, standard deviation 13.3 points for this model is identical to the RMSE spilled alone.5 In addition, despite the differences in the fcc structure and TradeSports Vegas markets, both perform the same. Next, we go beyond the RMSE into account separately for calibration and discrimination. F igure 4 shows the distribution of the total predicted and empirically observed results of six forecasting methods where prediction binned by 5% intervals, and each area of ​​the circle represents the number of predictions in the corresponding probability range.

A s should be clear from the figure, all methods allow the prediction that lie approximately on the diagonal. All methods are therefore reasonably well calibrated—predicted 5Cross-validation—also known as rotation estimation— protects against overfitting the model to the data. E vents are first partitioned into k = 5 subsets of approximately equal size, and then predictions are made for events in each of the k subsets via a model trained on the remaining k βˆ’ 1 subsets. probabilities agree with observed probabilities within any given bin—however, they differ in their ability to discriminate.

M ost notably, whereas the baseline model includes all events in the same bin, thereby effectively treating high and low probability events as indistinguishable, other methods distinguish between empirically likely and unlikely events, as indicated graphically by the dispersion of bins along the diagonal. T able 1 confirms these visual impressions, quantifying the calibration and discrimination of each method, and also reveals two main findings. F igure 3: Yearly performance of the Vegas market and the win-loss model in predicting final point differences (i.e., home team score minus away team score) in NFL games. Table 1: Calibration error, discrimination, and RMSE for several methods in predicting the probability the home team wins in NFL games. First, two of the six methods are inferior to others in one dimension or the other: the filtered polls discriminate good, but not as calibrated as other methods; while the base model is well calibrated, but does not discriminate. Secondly, the methods of the six four-Vegas markets TradeSports, Probability sports, and win-loss model are comparable both in terms of calibration and discrimination. T.

his poor discrimination of the base model is just a small fine in terms of RMSE due in part to the peculiarities of the NFL (e. g, salary caps), which ensure that the majority of games played between closely matched teams, and therefore decided to probably close to 50%. In other words, although the base model does not work bad for highand low-probability events, the relative rarity of such events mean that it does not break much for these failures. T herefore, we can suspect that in areas (such as policy analysis), where the events are not intended to coin tosses, and where possible, predictions of greatest interest can be for extreme probability event, less discriminatory methods will perform respectively worse than they do here. T o test this idea, we recalculated RMS prediction methods exclusively for six unbalanced pairings between «winning» teams (who have won at least nine of their last 16 games) and «losers» commands (lost at least 9 out of 16). T his subset includes 37% of the data.

A s expected, the basic model is made worse in these games (RMSE increased from 14.5 to 15.1 points), but the RMSE of TradeSports and Las Vegas markets, the probability of the sport, and a win-win model of the losses were roughly unchanged, 13.0 13.2, 13.4, and 13.5, respectively. T hus, even for the more these extremes, we find prediction markets again have only a small advantage over conventional methods of forecasting5.2 BaseballBasketball is widely popular around countries such as the US and Japan and some parts of Eastern Europe and Australia as well.

B asketball consists of two teams with six players who are trying to put the ball in the opponent’s basket. P layers who practice the basket quite higher than were found in any other sport. B asketball has gained so much popularity with its collegial sports like baseball and rugby, that the people of these countries like to bet and predict the fixtures using a number basketball software. B.

asketball played in two levels, Pro Basketball and College Basketball. B oth are equally popular among fans and supporters, and they like to predict on top expert pro basketball picks and college basketball forecasts. N umerous websites like Pdssports.com bettingexpert.com and offers a sports handicapping service, sports handicapping sites, expert sports picks, sports betting systems, winning sports betting in the network to supporters through the Internet. E xpert pro basketball picks and college basketball picks are available at cheap and affordable packages on these sites, which have been duly filed with listing a number of previous games databases and analysis.

B asketball software or Handicapping Software is also available that provides the ability to forecast the game, and the nominal capacity, evaluate the effectiveness of the line, a sequence of offensive, defensive, performance and valuation. Sports betting systems and strategies are also given as relevant modules over these sites for the beginners. T he sites provides near accurate analysis and intelligence about the future games to be played. N.

ot only Basketball but also Baseball and American football predictions and picks are available in these sites at affordable rates for every pocket. Forecasts for the NBA basketball are popular enough BETTER field activities. U sually, bookmakers offer a rich painting for the match, the most diverse statistics widely available (both teams and players), broadcast matches across the globe. I n particular, the Russian rights to broadcast NBA games belong ViasatSport television. Immediately I must say that the «professional» predictions for basketball, which offer to purchase numerous Capper, in most cases, is as reliable as your own will. If you do decide that someone else’s opinion in betting on basketball is more faithful than your own, and trust the experience of forecasters, we recommend that you use exchanges forecasts (not to be confused with the exchange rate). O.

f the benefits of forecasts, stock exchanges, compared with home-grown"professionals"kapperskogo cases, recording its uncontrolled flow of words in deplorable web-camera, the following advantages can be identified: Honesty approach to the process. A s a general rule, to declare himself a professional Capper and demand respect to these resources — a bad taste. M.

oreover, the players on the stock exchange forecasts are generally honest and do not introduce their clients to confusion by claiming that their strategy of win-win (after all, it is easy to check).Competition among forecasters. I n order to climb to the top sheet (and get more customers) simply state that you are a professional forecaster, will not be enough. F orecasters have to prove their worth real result. Honest statistics.

F orecasts Exchange itself is interested in that forecasters do not «cheat» their stats in basketball forecasts. F or this purpose there are effective control tools to stop all the possible attempts of fraud by forecasters. The relatively low price tag of forecasts. C ompared to the «professional» forecasters, trading right and left «glands» and «dogovornyakami», the participants exchange forecasts put quite a democratic price tag for its forecasts for basketball. For those who want to save their capital, we have prepared a selection of resources where you can get ready to predictions of future outcomes.Vitibet.com.Besplatno betting on basketball is available on the websites a la Vitibet.com and similar projects. O.

n the resource regularly laid out calculations of the probability of basketball betting outcomes (and forecasts for many other sports). I t should be understood that the proposed table with the outcomes — is not a «piece of iron» and not even ready to forecasts basketball games. I n fact Vitibetassessment of the likelihood of neural network based on statistical data.Typersi.com — online tournament Capper, sponsored by several bookmakers. F.

or rates and results of the contest participants bet-you can watch for free. T he site presents statistics of participants, focusing on that, you can make your own opinion about the accuracy and effectiveness of their forecasts for basketball. T his tournament is unique and has a lot of unique in the Internet. F.

ind online tournament for players to bet on sports, not so difficult. T he main disadvantage of this kind of sources are limited terms of tournaments. Y ou either have to immediately choose to guide long-term rates in the basketball tournaments, or be prepared to regularly find new ones. Championship komChampionat.com — fairly well-known and easy to use sports information resource, where under the heading «Snatch» journalists portal choose the most interesting bets each game round NBA league and share their opinions about them.

H ere you do not just dry numbers, but, at least, an interesting author’s opinion about the basketball events and entertaining literary passages. F ans bet on basketball and learn the latest news from the NBA championship, itis strongly recommended. Features betting prediction Basketball: what should be taken into account? basketball from denegItak that need to be taken into account, making predictions for basketball games? O n Web sites, many «professional» forecasters really not stated what they are guided in their forecasts, in addition to many years of experience. S ome information resources, dedicated to sports betting, it is advised to pay attention to factors such as: changes in the composition, the injury, the form of players, the motivation, the statistics of the previous games. Let us, in order: Injuries and changes in the composition.

O bviously the lack of well-known key and just important players in basketball has always taken into account and incorporated in the bookmaker offers you a rate. Y ou can only try to subjectively assess the degree of significance of the lack of player on the court. H ow your score will be true in every case, more will depend more on chance (or probability theory, if it so wish), however, appreciate the importance of injury better than the bookmaker’s real. Unseen circumstances. I.

t is worth to highlight the force majeure and unexpected news of the absence of the leaders of which have not been taken into account while setting the coefficient of the bookmaker. H ere they can be a decisive factor, and most importantly — not medlit. travma Carmelo AnthonyForm leaders. I f you consider yourself sufficiently knowledgeable in order to make your judgment on the form of the players found in the monetary expression of your bets — it’s wonderful. H owever, as practice shows, to objectively assess the current form of NBA players and its influence on the course of the match a man who does not have expert knowledge, it is very difficult.

I n addition, it is no secret that the anti-doping control in NBAformality than a real phenomenon. S o somewhere sunk fatigue NBA player past days should not be surprising.Psychology.

T ypical judgments «they lost the last game, so should: assemble / upset» or «they won the last game, so should be on the rise / relax» is highly subjective and is usually unfounded. S uch statements are good for professional journalists doing previews upcoming basketball games and looking for «spice». For a person making predictions, it is just a game of toss, which is better to get rid of and save their time.Statistics.

W here do without statistics today. H owever, all that you know (victories, defeats, basketball team points) — all this is available and the bookmaker.

M oreover, all this is taken into account by the bookmaker. I n this connection, it is recommended to pay special attention to changes in bookmaker quotes causes preference deliver. T hose.

— T rack betting odds changes taking place just before the match. T hey can change the objective assessment of the forces and create favorable conditions for your bet. I t is also recommended to use the method of calculating the rate of the Kelly index. Although we have considered a number of performance measures, it is possible that football remains a special case even in the domain of sports in that outcomes are dominated by hard to anticipant events—a hail Mary pass in the final minutes, for example, or an intercepted ball against the flow of play—for which there is relatively little real information on which to base sophisticated predictions. In addition to football, therefore, we consider Major League Baseball (MLB)—a sport for which very large amounts of data are collected, and where an entire field, sabermetrics, has been developed along with its own journal, the Baseball Research Journal, specifically for the purpose of analyzing performance statistics. I n light of this considerable devotion to statistical models and prediction, one might assume that expert observers, and hence prediction markets, would outperform simplistic models by incorporating game-specific variables like pitching rotation, the recent batting performance of individual players, and so on.

A s described belowFigure 4: Distribution of predicted and empirical probability estimates for the home team winning in NFL games. T he area of ​​each circle represents the number of predictions in the corresponding probability range. Nevertheless, we see that baseball markets have only a slight advantage compared to alternative prediction tools. W.

e compare the performance at Vegas market to the base line and win loss models for 19.633 Major League Baseball (MLB) games for seven years, from 1999 to 2006, where the two models in the same way as for football were built. I n particular, the base model ignores all the game specific information, always predict hosts won with a probability of 0.54 and historical winning percentage of home teams in baseball. Accordingly, the model is a win-win for the loss of baseball was identical to the form that is used for football predictions: when A and B teams play each other on a home field A, the probability A wins is estimated at B + (RARB) / 2, where L = 0 54 is a basic probability of winning the home team, an RA percentage of team games and won 162 of his last match-ups (number of regular-season games each team plays a year), and RB is an analogue for the team B. percentage terms of three performance indicators introduced above RMS, calibration and discrimination, we are once again, that the model is a win-win loss stands at the same level of the market (Table 2, 5 and 6) to find. I.

n particular, the market and the win-loss model both have an RMSE of Table 2: Calibration error, discrimination, and RMSE for several methods in predicting the probability the home team wins in MLB games. 0.

50.6 Furthermore, all three methods are well calibrated, with calibration errors of 0.02 for the market and the winloss model, and 0.01 for the baseline model. F inally, although the shortcomings of the baseline model are apparent from its inability to discriminate between high and low probability events, the market and win-loss model remain comparable by this measure as well, with discrimination 0.09 and 0.07, respectivelyConclusionsOne of the most important tasks of forecasting sporting achievementsIt is to increase the accuracy of calculations. A s is known, the accuracy of the forecast depends on the length of time series, and the period of retrospection period anticipation, forecasting method, and other factors [2].

G eneral opinion on how these factors affect the efficiency of higher prediction results, no. F or example, some scientists believe that the longer the period of history (20−30 years), the more reliable the forecast. O thers believe that the outlook for the future is better to include the next two to three years, and therefore forecasting time series does not need a long duration. T.

hirdThey find it impossible to describe the long time series of a single equation, and assume expanded into separate simpler parts with a duration of 4−8 years. T his gives allegedly possible to describe those using linear equations and parabolas of second and third orders. There are currently poorly understood and the question of the optimum ratio of the length of the period of retrospection and anticipation during the period predicting the results in a particular sport. N.

ote that in general prognostics experts suggest taking them equal to orchoose the term period, equal to 1/3 of the number of reported data [2]. S ometimes out that forecasting period should not exceed the length of the time series and time series must be at least 10 years. In particular, the Table. 8 — 9 shows that in spite of the projected long enough period (up to 30 years), mean forecast error, in most cases, both males and females is low. N oteworthy is the fact that the reduction of the forecast period and the increase in the period retrospection it decreases. The accuracy of forecasts obtained by expert assessments and computer modeling, higher than the accuracy of linear extrapolation of time series. The accuracy of forecasts developed by extrapolation method, increases with decreasing the lead time of the forecast. These tables indicate that the value of the average error in the forecast track and field disciplines in women than in men. Thus, on the basis of the materials discussed in this chapterWe can draw the following conclusions:

1. Prediction of sports the highest achievements in sports, with the metric system of measurement, despite its apparent simplicity, is a very complex and poorly understood problem, although she devoted many publications. The reason — the high level of uncertainty in the dynamics of growth results in the individual sports disciplines and the conditionality of development by many factors, the impact of which is not always possible to establish and formalize.

2. In assessing the overall positive forecasts of world achievements and results of winners and prize-winners of the Olympic Games, made in the last three decades, we have to note that in most cases they were unreal. Their authors were unable to avoid the blunders and miscalculations in anticipation of higher future sports achievements.

3. Conduct us a posteriori verification of forecasts shown their accuracy, ie, the degree of compliance with the predicted and actual results will depend on the specific sport discipline and the nature of the dynamics of growth of world and Olympic records (straight, -foot, curved), gender of athletes, the duration of the period of prehistory (Flashbacks), based on which the forecast is made, and the forecast period — the period of pre-emption, the chosen methods of forecasting and availabilityInformation about accelerating and inhibiting factors affecting the development of the sport of the highest achievementsReferences.

ΠŸΠΎΠΊΠ°Π·Π°Ρ‚ΡŒ вСсь тСкст

Бписок Π»ΠΈΡ‚Π΅Ρ€Π°Ρ‚ΡƒΡ€Ρ‹

  1. [1] K. J. Arrow, R. Forsythe, M. Gorham, R. Hahn, R. Hanson, J. O. Ledyard, S. Levmore, R. Litan, P. Milgrom, F. D. Nelson, G. R. Neumann, M. Ottaviani, T. C. Schelling, R. J. Shiller, V. L. Smith, E. Snowberg, C. R. Sunstein, P. C. Tetlock, P. E. Tetlock, H. R. Varian, J. Wolfers, and E. Zitzewitz. The promise of prediction markets. Science, 320(5878):877−878, 2008.
  2. [2] J. E. Berg, R. Forsythe, F. D. Nelson, and T. A. Rietz. Results from a dozen years of election futures markets research. In C. R. Plott and V. Smith, editors, Handbook of Experimental Economics Results, Volume 1, pages 742−751. North Holland, 2008.
  3. [3] J. E. Berg and T. A. Rietz. Prediction markets as decision support systems. Information Systems Frontiers, 5(1):79−93, 2003.
  4. [4] Center for Gaming Research, University of Nevada, Las Vegas. 2008 Nevada gaming statewide revenue breakdown.
  5. [5] Y. Chen, C. Chu, T. Mullen, and D. Pennock. Information markets vs. opinion pools: An empirical comparison. In Proceedings of the 6th ACM conference on Electronic commerce, page 67. ACM, 2005.
  6. [6] Y. Chen and A. M. Kwasnica. Security design and information aggregation in markets, 2006.
  7. [7] J. D. Christiansen. Prediction markets: Practical experiments in small markets and behaviours observed. Journal of Prediction Markets, 1(1), 2006.
  8. [8] R. Clemen. Combining forecasts: A review and annotated bibliography. International Journal of Forecasting, 5(4):559−583, 1989.
  9. [9] V. Dani, O. Madani, D. Pennock, S. Sanghai, and B. Galebach. An empirical comparison of algorithms for aggregating expert predictions. In Proceedings of the Conference on Uncertainty in Artificial Intelligence (UAI). Citeseer, 2006.
  10. [10] R. S. Erikson and C. Wlezien. Are political markets really superior to polls as election predictors? Public Opinion Quarterly, 72(2):190−215, 2008.
  11. [11] E. Fama. The behavior of stock-market prices. Journal of business, 38(1):34, 1965.
  12. [12] M. Ferrari and A. Rudd. Investing in movies. Journal of Asset Management, 9(1):22−40, 2008.
  13. [13] R. Forsythe and R. Lundholm. Information aggregation in an experimental market. Econometrica, 58(2):309−347, 1990.
  14. [14] R. Forsythe, F. D. Nelson, and G. R. Neumann. Anatomy of an experimental political stock market. American Economic Review, 82(5):1142−1161, 1992.
  15. [15] R. Forsythe, T. A. Rietz, and T. W. Ross. Wishes, expectations, and actions: A survey on price formation in election stock markets. Journal of Economic Behavior & Organization, 39:83−110, 1999.
  16. [16] A. Gelman, J. Carlin, H. Stern, and D. Rubin. Bayesian data analysis. Chapman & Hall, 2003.
  17. [17] S. Goel, J. Hofman, S. Lahaie, D. M. Pennock, and D. J. Watts. What can search predict? Technical Report.
  18. [18] A. Graefe and J. S. Armstrong. Predicting elections from the most important issue facing the country, 2009.
  19. [19] R. W. Hahn and P. C. Tetlock, editors. Information Markets: A New Way of Making Decisions. AEI-Brookings Press, 2006.
  20. [20] R. Hanson. Decision markets. IEEE Intelligent Systems, 14(3):16−19, 1999.
  21. [21] R. Hanson and R. Oprea. Manipulators increase information market accuracy, 2004.
  22. [22] R. Hanson, R. Oprea, and D. Porter. Information aggregation and manipulation in an experimental market. Journal of Economic Behavior & Organization, 60(4):449−459, 2006.
  23. [23] F. A. Hayek. The use of knowledge in society. American Economic Review, 35(4):519−530, 1945.
  24. [24] P. Healy, J. Ledyard, S. Linardi, and R.J.Lowery. Prediction market alternatives for complex environments. In Conference on Auctions, Market Mechanisms and Their Applications, 2009.
  25. [25] J. Howe. Crowdsourcing: Why the Power of the Crowd Is Driving the Future of Business. Crown Business, New York, 2008.
  26. [26] J. C. Jackwerth and M. Rubenstein. Recovering probability distributions from options prices. Journal of Finance, 51(5):1611−1631, 1996.
  27. [27] F. Kleeman, G. G. Voss, and K. Rieder. Un (der)paid innovators: The commercial utilization of consumer work through crowdsourcing. Science, Technology & Innovation Studies, 4(1):5−26, 2008.
  28. [28] J. Ledyard, R. Hanson, and T. Ishikida. An experimental test of combinatorial information markets. Journal of Economic Behavior & Organization, 69(2):182−189, 2009.
  29. [29] S. Makridakis and M. Hibon. The M3-Competition: results, conclusions and implications. International Journal of Forecasting, 16:451−476, 2000.
  30. [30] S. Makridakis, M. Hibon, and C. Moser. Accuracy of forecasting: An empirical investigation. Journal of the Royal Statistical Society. Series A, 142(2):97−145, 1979.
  31. [31] S. Makridakis, R. M. Hogarth, and A. Gaba. Forecasting and uncertainty in the economic and business world. International Journal of Forecasting, In press, 2009.
  32. [32] J. Muth. Rational expectations and the theory of price movements. Econometrica, 29(3):315−335, 1961.
  33. [33] K. Oliven and T. A. Rietz. Suckers are born, but markets are made: Individual rationality, arbitrage and market efficiency on an electronic futures market. Management Science, 50(3):336−351, 2004.
  34. [34] D. M. Pennock, S. Debnath, E. J. Glover, and C. L. Giles. Modeling information incorporation in markets, with application to detecting and explaining events. In Proceedings of the Eighteenth Conference on Uncertainty in Artificial Intelligence, pages 405−413, Edmonton, CA, 2002. Association for Uncertainty in Artificial Intelligence.
  35. [35] C. R. Plott and S. Sunder. Efficiency of experimental security markets with insider information: An application of rational-expectations models. Journal of Political Economy, 90(4):663−698, 1982.
  36. [36] C. R. Plott and S. Sunder. Rational expectations and the aggregation of diverse information in laboratory security markets. Econometrica, 56(5):1085−1118, 1988.
  37. [37] C. Polk, R. Hanson, J. Ledyard, and T. Ishikida. Policy analysis market: An electronic commerce application of a combinatorial information market., 2003. [38] P. W. Rhode and K. S. Strumpf. Manipulating political stock markets: A field experiment and a century of observational data, 2006.
  38. [39] R. Roll. Orange juice and weather. American Economic Review, 74(5):861−880, 1984.
  39. [40] R. N. Rosett. Gambling and rationality. Journal of Political Economy, 73(6):595−607, 1965.
Π—Π°ΠΏΠΎΠ»Π½ΠΈΡ‚ΡŒ Ρ„ΠΎΡ€ΠΌΡƒ Ρ‚Π΅ΠΊΡƒΡ‰Π΅ΠΉ Ρ€Π°Π±ΠΎΡ‚ΠΎΠΉ
ΠšΡƒΠΏΠΈΡ‚ΡŒ Π³ΠΎΡ‚ΠΎΠ²ΡƒΡŽ Ρ€Π°Π±ΠΎΡ‚Ρƒ

Π˜Π›Π˜