Close predictions for 12/12/2012

[datumiq]

Close predictions for 12/12/2012
Released 9:45am at DatumIQ.com
Back tested over the last 10 years

82% chance ACN closes below 72.28
86% chance ADP closes below 58.74
84% chance AFL closes below 54.96
86% chance AXP closes above 57.12
85% chance BAC closes above 10.41
80% chance C closes above 36.54
80% chance CNQ closes below 29.19
81% chance COP closes above 58.18
84% chance CTSH closes below 74.14
84% chance DIS closes below 50.12
81% chance EMC closes below 25.27
85% chance EPD closes below 50.33
82% chance F closes above 11.44
79% chance FCX closes above 31.65
82% chance GG closes above 36.87
81% chance GOOG closes below 709.66
83% chance GS closes below 121.21
81% chance HOG closes below 49.08
83% chance INTC closes below 21.48
84% chance IWM closes below 84.39
90% chance JNJ closes above 71.02
80% chance LTM closes below 50.02
82% chance MCD closes below 90.18
84% chance MMM closes below 95.23
83% chance ORCL closes below 32.53
83% chance PM closes above 87.44
82% chance QCOM closes below 65.16
83% chance QQQ closes below 67.04
83% chance SPG closes above 153.59
88% chance SPY closes below 145.10
81% chance SU closes below 32.72
85% chance SYK closes below 56.65
84% chance T closes below 34.77
85% chance TEVA closes above 40.05
85% chance WMT closes above 68.66

 

[Shakone]

What do you base that on? If the 90% chance one turns out to be wrong, how confident are you the other %'s are still 'accurate'?

 

[datumiq]

What do you base that on? If the 90% chance one turns out to be wrong, how confident are you the other %'s are still 'accurate'?

Each probability for a stock is independent of a probability of another stock.
The probability is based off applying my proprietary algorithm on the last 10 years of a stock (approximately 2500 trading days). On my site when you click one of the tickers, the back tests are shown for the last 10 years to show the accuracy of the algorithm.

 

[wackypete2]

Of the ones that I checked they all opened this morning well in the direction that you are predicting. Seems natural to me that the percentages would be high in that case. There'a a better chance that if I bought/sold on open based on the direction you suggest I could lose but your "prediction(s)" would still be correct.

Show me something less obvious.

Peter

 

[Shakone]

Each probability for a stock is independent of a probability of another stock.
The probability is based off applying my proprietary algorithm on the last 10 years of a stock (approximately 2500 trading days). On my site when you click one of the tickers, the back tests are shown for the last 10 years to show the accuracy of the algorithm.

How do you know the probability is independent?

 

[datumiq]

Of the ones that I checked they all opened this morning well in the direction that you are predicting. Seems natural to me that the percentages would be high in that case. There'a a better chance that if I bought/sold on open based on the direction you suggest I could lose but your "prediction(s)" would still be correct.

Show me something less obvious.

Peter

That's a good point. What I can do is before the open (9:30am) I can give the probability the close will be plus/minus a certain amount.
For example if GE closes at 21.50 on 12/12. Then on 12/12 in the evening I can release the closing prediction for GE on 12/13, such as "GE closes below Open Price + 0.50, 85% probability ". So if whatever the open price is, just add 50 cents to it and that is the ceiling I am predicting.

I appreciate your feedback.

 

[datumiq]

How do you know the probability is independent?

That's a good question. The probability of a stock is derived off patterns only within itself. So I do not look at the movements of other stocks when looking at one.

 

[Shakone]

That's a good question. The probability of a stock is derived off patterns only within itself. So I do not look at the movements of other stocks when looking at one.

So the probability is not independent then...

How do you trade on this info?

 

[datumiq]

So the probability is not independent then...

I'm sorry I mis-understood your question. I thought you were asking if it is dependent on any other things other than itself. In terms of the history of itself, you are correct it is based off it's own history.

 

[datumiq]

How do you trade on this info?

As mentioned in the previous message

What I can do is before the open (9:30am) I can give the probability the close will be plus/minus a certain amount.
For example if GE closes at 21.50 on 12/12. Then on 12/12 in the evening I can release the closing prediction for GE on 12/13, such as "GE closes below Open Price + 0.50, 85% probability ". So if whatever the open price is, just add 50 cents to it and that is the ceiling I am predicting.

So in terms of trading if the prediction is GE will close above it's open price with 85% probability. Then if GE at anytime falls under the open price, then it can be considered to go long for the day since 85% chance it will close above open.

 

[Shakone]

As mentioned in the previous message

What I can do is before the open (9:30am) I can give the probability the close will be plus/minus a certain amount.
For example if GE closes at 21.50 on 12/12. Then on 12/12 in the evening I can release the closing prediction for GE on 12/13, such as "GE closes below Open Price + 0.50, 85% probability ". So if whatever the open price is, just add 50 cents to it and that is the ceiling I am predicting.

So in terms of trading if the prediction is GE will close above it's open price with 85% probability. Then if GE at anytime falls under the open price, then it can be considered to go long for the day since 85% chance it will close above open.

Hmm, in terms of probability this is a mistake though isn't it.

You can look at historical prices and calculate a range it is likely to fall in. However, GIVEN that it drops below that price, the probability it will close above is a very different probability than the one which states it will close above the value. Make sense?

 

[datumiq]

Hmm, in terms of probability this is a mistake though isn't it.

You can look at historical prices and calculate a range it is likely to fall in. However, GIVEN that it drops below that price, the probability it will close above is a very different probability than the one which states it will close above the value. Make sense?

Good analysis, however the algorithm does not take that into account since majority of cases the daily low is lower than Open and daily high is higher than Open. There are a few cases where Open is the low or high. But for a trading strategy if the prediction is price will close above open and price during the day goes below open then consider buying long for that day. I do understand what you're saying on the intraday movement, I will see if there are any decent patterns with that included.

 

[Shakone]

Not sure we understand eachother.

You gave some data, say 80% google will close below a price X. You used historical data for this.

Then I asked how you used this info to trade. I got the impression you were suggesting, that if Google rose above X at any time in the day, then you could short google, because it had an 80% chance of closing below X. This conclusion would be false. But I may have misunderstood how you are suggesting to trade on it.

So given say '81% chance GOOG closes below 709.66', how would you suggest trading on that information?

 

[BeginnerJoe]

So given say '81% chance GOOG closes below 709.66', how would you suggest trading on that information?

I'd imagine you would sell during the day if the price goes above 709.66 and buy back at close for profit.

All these imaginary/demo pips/points are really worthless. Let's see some real pips/points.

 

[Shakone]

I'd imagine you would sell during the day if the price goes above 709.66.

All these imaginary/demo pips/points are really worthless. Let's see some real pips/points.

Yes, this is what I thought he was suggesting. But this is a probability mistake.

 

[datumiq]

Yes, this is what I thought he was suggesting. But this is a probability mistake.

For the trade 2 example would be as mentioned by "BeginnerJoe - I'd imagine you would sell during the day if the price goes above 709.66 and buy back at close for profit."

But what I'll do to keep it simple and valuable, is the evening before the trading day I will release a simple prediction if it exists for a stock:
Example:
"75% chance GOOG closes higher than Open price"

So for trade, if GOOG falls under the Open Price, then the trade would be to go long for the close since there is a 75% chance of being higher than the Open.

I appreciate all of the feedback to improve/clarify my data mining software.

 

[datumiq]

Hey guys as mentioned I put out the 12/14/2012 predictions out the day before.
I created a new thread for it.

http://www.trade2win.com/boards/technical-analysis/163328-close-predictions-12-14-2012-a.html

-Rob

 

[Shakone]

I know, but it still suffers from the same problem. There's no trading strategy here, and even IF we accept the probabilities as true, the strategy you suggest would still be wrong probabilistically.

 

[datumiq]

I know, but it still suffers from the same problem. There's no trading strategy here, and even IF we accept the probabilities as true, the strategy you suggest would still be wrong probabilistically.

Just so I fully understand. Can you give me an example of a few predictions which would be valuable for trading.

 

[Shakone]

If you have stats related to where the price will go from the open, then those stats are fairly meaningless later on when price has already moved. e.g. EURUSD 80% chance it won't move upmore than 100 pips from open (say). If price has moved up 80 pips, then that 80% figure is now redundant, and so of no use to any trading method. It is no longer a relevant probability. You would need to calculate the probability of it not going up more than 100 pips, conditional on the fact it's already moved up 80 pips. This is not a probability you're quoting.

So the strategy suggested in the other thread:
"A possible trading strategy is if the stock price does not pass the Open price + the price specified, then consider going long until it hits the minimum high predicted," again doesn't make sense.

e.g. you said "GOOG 77% chance Day High is at least 2.86 greater than open price". Your suggested strategy then implies, if google drops from open, and continues dropping for whatever period of time,1 hour, 2 hours, 4 hours etc, it would still make sense to buy because you think the price has a good chance (77%) of having a higher high, 2.86 above the open. But again, it doesn't. You don't understand the probability here. It is no longer 77% chance of going higher than 2.86+open. Try to get this point clear.

So if you're intending to use this for trading, you have probability that's relevant from the open (and only the open), and therefore you need a strategy which makes a decision at the open.

So if you agree with all that, then for trading purposes, the next step. It's not enough to suggest Goog high is 77% chance 2.86 more than the open. Because suppose I go long with profit target at 2.86 higher than open. I would be concerned of
1)How large a stop will I need to not get stopped out on those 77% times
2)How much could I lose on the other 23% of times.

So really you need a sort of:
with 77% probability it will go higher than X and not go lower than Y, or goes higher than X before going lower than Y. Then someone could make a decision about whether it is worthwhile