can you predict the *PAST* from now?

a crazy thought crossed my head before going to sleep!

all technical analysis is centered around predicting the future move from current signals...

but what about the reverse? is the inverse possible?

if I gave you this month chart... is it possible to predict what would have happened in the previous month (ofcourse with out seeing the previous months charts) ?

If this is possible I think we can objectively evaluate the success of indicators... without guessing or blaming randomness !

onlinegtrash said:

a crazy thought crossed my head before going to sleep!

all technical analysis is centered around predicting the future move from current signals...

but what about the reverse? is the inverse possible?

if I gave you this month chart... is it possible to predict what would have happened in the previous month (ofcourse with out seeing the previous months charts) ?

If this is possible I think we can objectively evaluate the success of indicators... without guessing or blaming randomness !

Click to expand...

I believe even the 'inverse' is not possible.
but why the emphasis on 'prediction'??

regards

crazy one ... but still worth to think i guess..

yeah 'prediction' won't help much... but wondering about this inverse...

for instance... a doctor can easily identify the smoking behavior of person's past with his latest liver report... we don't trust the doctor if he can't do this.

Similarly if we can't predict backwards... I think most TA should be bundled under 50-50 random fluke... but I do believe TA helps to identify market structures such as support resistance, channels etc... but hardly direction from there...

It is the same. Imagine your chart reversed from right to left. And start drawing candles from now to past . Then all indicator/chart studies/patterns will work as it is. The indicators are not depended on time.

But the fact remains that what we are trying to get is probability there is nothing called 100% and the same should hold true when trying to predict the past from present.

It should be another way to backtest a strategy. May be we can call it reverse back testing :! but I don't see it will add any value.

altho i ve never reverse back tested... anything...
but i have a feeling right-to-left-back testing (r2l-bt) will completely give different results from the traditional left to right back testing (l2r-bt) !
bcoz the entry points r2l-bt should be exit points in l2r-bt to have the similar results... which I think can't happen...

definitely FNO segments can't have such symmetrical results but am not sure this can happen to cash segment stocks either... stocks fall crashes are always sharper and quicker than bull markets.... in r2l-bt... you will be seeing overnight successes... and a simple algo will have much easier time from bailing out losses in r2l-bt and super easier time in making money with overnight reverse crashes!

so my hypothesis is... r2l-bt will be always more profitable than l2r-bt...not bcoz of look-ahead effect...but market structure has something todo with it!

i might be wrong...but guess it will be fun mental exercise to understand 'time and structure' in markets!

I don't agree with the highlighted part.
Although I agree r2l-bt backtesting result will be different than l2r-bt as the data set is different(becuase of inverse) but we can not say which one will more or less profitable than the other

if u can predict future in a chart u can tell its past also. but of no use why wasting time over it all the money lies in predictinig future.

Its simple ! Hold d chart image upside down .