Large Trades Distort Prediction Accuracy replied by Roger @ Wed, 25 Jun 2008 18:40:47 -0000

Roger — Wed, 25 Jun 2008 18:40:47 -0000

Hi Bill
It seems like my point wasn’t clear enough.

In Inkling’s design of a prediction market, the size of a single position can determine the prediction percentage of the total position holders in a market. This means, if a position holder has a large account, they can easily distort the predictive range of agreement, or disagreement.

This isn’t reality in the real world, and it isn’t reflective of crowd estimation research. For example, when I place a trade for 5,000 contracts of Corn, or 3,000 contracts of Crude Oil, if my broker is careful and does it as volume appears in the market, my entry price into the market can end up being near where it was just before I made entry. Now, if other people noticed what I’m doing, and that happens sometimes, they can try to run my stops, or get on in the same direction.

If a lot of people get on in the same direction, this moves the market in the direction of my trade. However, if I’m the only one who gets in, the market will hardly notice my position and the price won’t change much until the volume of trading begins to move either against or in favor of my position. This means that for the market price to change, there needs to be broader market participation, not large size, in order for the price to be significantly impacted.

Inkling’s approach views the size of a large position as broad market participation, instead of a large trader with a big paw. It also means that if you have a large paw, your prediction bias is more likely to be correct than the average of all previous traders. This is a silly concept because it believes “might it right”, through its interpretation of position size. Reality shows that good predictions come from the average results of a crowd, not from the individuals in the crowd who have big pockets.

It isn’t hard to see why Inking might have designed the process to work this way because it simplifies the process of making prediction value adjustments, and because it creates a game that looks like a real world market. However, my argument is that they should have gone in a different direction. For example, let’s assume a large player with 2m Inklings decides to risk 500 contracts in a market, but that position size in comparison to his available Inklings calculates to less than 0.025% of his Inkling equity. Is this large trader thinking his prediction decision is rock solid? It doesn’t seem like it when you compare it to someone with an account of 5,000 Inklings who thinks they know something about the market, so they decide to risk 500 Inklings, or 10% of their account balance in that market. Should the not-so-sure large trader’s position size influence the price more than the person who believes they know what is going to happen, and puts 10%, or a larger piece of themselves at risk? I don’t think it is reality.

My view, if we are to maintain the market model appearance, is to have the risk, in terms of the exposure to a trader’s account be the most influential reflection of his belief be the impacting element when it comes time for Inkling’s Market Maker to adjust the market’s prediction range values. Let’s put this is terms of how each of us would react to two different sets of facts. A new expert who doesn’t know much says, he thinks, “your chances are good, but don’t hold me to a certainty.” However, the next expert says, “controlled research shows that if this approach is used the chance for success is going to be better than 40%, and I’m willing to put my reputation on the line that the information is accurate.”

Would you believe the person who doesn’t want to risk anything more than your frustration, or the person who is willing to take a serious reputation hit, and has supporting data that his answer is correct? My instincts have me going with the person who did the work, stated his certainty and is willing to stand behind what they are saying.

Another example might be when we ask for an opinion and ask a group of people to give us their best guess on what might happen. As we go through the group we ask each person to give us a prediction, and also their belief in how strongly they feel about the certainty of that prediction. If all the people, but one said, their average belief indicates an outcome of a 30% chance for success, but one person demanded we write down the outcome had a 90% certainty and would not stop talking. What would you give the outcome to be from this sampling? Inkling’s calculation would let the stronger person have a large influence on the outcome.

If the Inkling process would look at what people are willing to risk as a percentage of what they have available, then they are getting a percentage number or value from each trader that can then be analyzed to create a prediction percent. This is in reality the information they are getting in the prediction markets that require people use actual cash to describe their belief level. By allowing people to think in terms of how strongly they believe in a prediction, versus the current Inkling paradigm of how much they have to bet, the prediction accuracy would be better and the percentages would move away from what people believe might be a reasonable gamble of where they can make a commitment, to what they think will happen.

Another area that distorts the process is how much the number of Inklings available can allow or restrict how many markets a participant can record their belief in the markets they find interesting. This issue is based in some of the above logic here and because of the market model being presented, but this posting is too long already.

As for the prediction amount being so dependant on large trade size, and allow large traders like yourself to take a large rake from better entry price into a market, I agree that would be a natural result of what I’m proposing here. Of course, that has a downside when the guess is wrong, but that is the beauty of making the process more participation and belief dependant.

When size is less important, accuracy and consistency increase. What we have now are large paws putting low risk positions in a lot of places distorting the reflection of what people think might happen, and doing this with a smaller sample of opinions than would result otherwise. Crowd size is an important aspect for finding the value of something with an opinion process.

Large Trades Distort Prediction Accuracy replied by wstritt @ Mon, 09 Jun 2008 16:08:51 -0000

wstritt — Mon, 09 Jun 2008 16:08:51 -0000

I’m not sure I understand your point. Of course large trades move the market in a significant way. I don’t understand, however, why that necessarily “distorts” prediction values.

You note that in the real world the impact of a clumsily executed large trade would revert to prior market levels whereas in the Inkling world it doesn’t. In the real world that is only true in the event the large trade is actually distorting (i.e. sets wrong price) as sellers will enter the market and drive down the price to its appropriate equilibrium. If the trade ultimately reflects new information, it will not revert. In either case, there is generally sufficient market liquidity on both sides of the question to set the “right” (one could quibble with that adjective) price, though clearly not in all cases (e.g. the current illiquidity in the mortgage backed markets).

In theory, if a large buy (or sale) in an Inkling market moves a price beyond what is reasonable, other traders are free to sell their positions or short the stock (or buy) and drive it back to where it belongs. Certainly both you and I have done so. Indeed, other traders should thank the distorter for creating opportunities/liquidity to take inklings off the table and/or earn outsized returns for the risk being taken. There are certainly a number of markets with a larger number of traders that appear to be reasonably efficient. If an Inkling market doesn’t revert, presumably other traders believe the the new prediction value is close enough to correct that there is no value to be had in taking a contrary view. In all of the above cases, there is no particular reason to think that the predictive value of one large trader (with conviction) is no better or worse that the predictive value of a bunch of smaller traders. Indeed, if the value predicted by one large trader doesn’t revert in the face of all traders presumably seeing it, the fact that everyone else doesn’t sell arguably implies that the other traders agree so the price does reflect the view of the many.

On the other hand, the alternative is that there is not enough liquidity to drive the price back to where it belongs. However that, presumably, is the distorter’s problem, not anyone elses (OK, maybe the market creator’s if they were using the results for something but they can read in to the lack of trading whatever they want). For what it is worth, I subscribe to that theory – there is not enough liquidity in many Inkling markets.

While I can’t speak to the trading strategies of other large balance traders, as one of them I can only tell you what I do. I almost never invest in more than 50 shares of anything unless I think that a stock is grossly over or under valued based on objective research. Often that means that the market isn’t well followed (no liquidity so minimal predictive value until I enter), or I happen to check in shortly after new markets open with default prices where outsized profits are available. Given that there are relatively easy inkles to be had with that strategy, there is little point in making large uninformed clumsy speculative investments. If anything, I’m guessing that my participation, and that of other large balance players, makes markets more efficient/accurate rather than distorted. In my view, if anything, there are not enough large traders to make every market efficient/accurate – take a look at my separate posts under Locks of the Week – opportunities that should not exist if the masses were paying attention and/or had plenty of liquidity.

Maybe your real complaint is that large balance traders find and exploit the lucrative investment opportunities before the small guys had a chance. Maybe, or maybe the large balance folks are more informed or interested in doing a little research. Last time I checked, the internet was available to every Inkling trader while new markets appear to be launched somewhat randomly. I will admit that large inkle balances probably provide more freedom to take positions in longer pay out markets which is otherwise a sub-optimal investment strategy for lower inkle balance players. I would submit that it makes those markets more predictive though.

To be honest, I’d love your solution to have prices automatically revert to pre-buy levels which would allow me to buy grossly under valued stocks over and over again.

By the way, my potentially “distortive” positions of 250 shares or more are listed below. Please let me know where you think I’ve grossly distorted the true probabilities aside from not buying or shorting enough (i.e. Yes I know the probability of Frei scoring the most points in Euro 2008 is 0, I just don’t have enough inkles to drive it to that level).

1824 long on when the 787 will be delivered
1554 long on the next NASA mission post shuttle retirement
1317 long on when 24 will premier
250 -1,000 shares short in the stocks of primary losers and other non-participants in the New Mexico election markets.
800 short on when the Pittsburgh Casino wil open
600 long on where Boston City Hall will be on 6/21
587 long on when Belgium will have its next astronaut
500 long that a Democrat will have the delegates necessary to win by 8/1 according to CNN & Fox
400 short that Arthur will be the strongest Hurricane of the season
400 short that Michael Strahan will play for someone other than the Giants
382 long that the Amazing Race 13 will premier in September
300 short that Wikipedia will hit 2.5MM English articles by 6/15
250 short in will McCain pick a VP today
250 short that Alexander Frei will score the most goals in Euro 2008
250 short that Rogers Clemens will be doing anything other than “something else” or “defending against perjury charges” on July 1
250 short that the US will have an Avian Flu outbreak before 7/1/2008
250 long that Hillary will speak at the Democratic convention
250 short of every stock in the Battlestar Galactica “Final Five” market except for the 4 already identified members

Large Trades Distort Prediction Accuracy replied by Roger @ Wed, 04 Jun 2008 17:04:37 -0000

Roger — Wed, 04 Jun 2008 17:04:37 -0000

Large Inkling Accounts distort prediction percent values, and thus accuracy, in a way that doesn’t reflect how physical markets (Currencies, Bonds, Grains, Energy, Stock, etc.) trade, or surveys would predict. When larger volume players enter a market by pushing a large size into a single transaction, the market will spike in the direction of the trade and stay there. This is no different than how the Inkling markets react, but what is different is that in the physical world markets, the prices revert to a value that is near to where it was prior to the clumsy way the position size was entered into the market. This is especially true since the electronic markets have now become the dominant way of trading in real world markets. If the Inkling markets were judged based upon a daily survey, the bias of the survey would not react to the large voice of a single person, but would count that opinion in comparison to how it favored the collective opinion’s bias.

In the case of how our Inkling markets react, the volume is interpreted as a larger more significant opinion of what will result, when in reality all that is being reflected is someone who is willing to bet a large size because they have a larger balance. There is no significant positive correlation between bet size and prediction accuracy, so trade-size shouldn’t be a controlling influence on prediction bias. Instead, a lot of the same opinions should be the dominant predictor and that only begins to have significance once a larger enough sample of opinions have been collected.

From watching how markets are playing out, larger size doesn’t improve the prediction, but it does push the prediction pricing around enough that it creates exit signals for prudent players, forcing them to leave the market because of the prediction’s price impact. It also distorts the market’s possible win/loss ratio enough to make the market unattractive to participation, or even to stay with the current position of the market. Taking new, or staying with current trades where the current Win/Loss ratio can cause a negative distortion to the trader’s positive expectancy percentage is a critical factor to avoid for a player/trader’s survival in the real world, and it is no different for a prediction player.

Instead of leaving the pricing so distorted and unrepresentative of accumulated opinions, it would be closer to reality that when a position size is applied to a market, that trade’s pricing should continue to be priced based upon how large the accumulated volume is impacting the current market’s prediction value, which is what happens now. However, and after the transaction is completed, the prediction price should then be allowed to return to a level near to where it was previously so that other participants in the market will understand the pricing spike was the result of a large volume single transaction of one player having an opinion, instead of a large shift in the opinion’s of the collective participants.

In simple terms, as more people accumulate volume on one side of a market, that should be the mechanism that moves the prediction in a direction, not the single size of a single bet size. To do otherwise changes the dynamics and reduces validity of the prediction markets. It also increases the risk of participating in a market because large volume trades are not viewed as aberrant price spikes.

I know that by pushing the prediction pricing around in Inkling’s current brutal fashion, people are avoiding participation in markets because the risk versus the reward of participating is too far out of line of what is acceptable. Low participation in the markets can only reduce the value of a prediction market, which doesn’t serve anyone’s best interest.

Please consider applying some average size weighting for adjusting the prediction pricing so that the collective opinions are more representative of how the market’s participants believe the question will be resolved.

Recent Posts in 'Large Trades Distort Prediction Accuracy ' | Beast

Large Trades Distort Prediction Accuracy replied by Roger @ Wed, 25 Jun 2008 18:40:47 -0000

Large Trades Distort Prediction Accuracy replied by wstritt @ Mon, 09 Jun 2008 16:08:51 -0000

Large Trades Distort Prediction Accuracy replied by Roger @ Wed, 04 Jun 2008 17:04:37 -0000