Comments on SquirrelInHell: Prediction Calibration - Doing It Right

Darn, that was quite decent app. But it stopped wo...

2019-11-09T12:08:37.803+00:00

Darn, that was quite decent app. But it stopped working complaining that it expired 2019-01-01.

And no source code are available to tinker with, e.g. recompile without meaningless date limit.

Here is a lesson and a plea for future generations: please make you apps open source, unless you have really strong reasons to keep you source code hidden.

Really nice. I will try out the app for sure (~99%...

2017-10-25T08:32:53.803+01:00

Really nice. I will try out the app for sure (~99% confidence) :)

OK, here's some pseudocode - I hope it is more...

2017-04-16T19:29:53.448+01:00

OK, here's some pseudocode - I hope it is more clear now!

// input:
// PREDS - an array that for each confidence level,
// contains the number of predicted outcomes at that level
int[1..99] PREDS := ...
int SUM_CORRECT := 0
int SUM_FAILED := 0
for i = 51 to 99
begin
SUM_CORRECT += PREDS[i] * 100 / i
SUM_FAILED += PREDS[100-i] * 100 / (100-i)
drawDataPoint("CORRECT", i, SUM_CORRECT)
drawDataPoint("FAILED", i, SUM_FAILED)
end

Could you post the actual algorithm, formally (I&#...

2017-04-16T17:46:28.092+01:00

Could you post the actual algorithm, formally (I'm not sure how to reconstruct the graph you provided from your description of it)?

Let me know how it goes, and if you have any sugge...

2017-02-16T20:17:19.926+00:00

Let me know how it goes, and if you have any suggestions on how to further improve the app.

@ open source: Not currently. If it happens, then I will still keep some of the additional/advanced features private, to have the option of creating a paid version of the app.

App seems really nice! I will give it a roll. If I...

2017-02-16T10:14:30.618+00:00

App seems really nice! I will give it a roll. If I find your approach really clearer than the usual one, I will start recommending it instead of my own Predict app http://lesswrong.com/r/discussion/lw/mmp/predict_log_your_predictions_app/
Do you plan on opening the source code?

As for the first request, acknowledged but I'm...

2017-02-01T05:34:20.865+00:00

As for the first request, acknowledged but I'm not promising anything.

As for the second, this is is already possible if you long tap a prediction and choose "Try Again".

Wow! That was fast! Some feedback on how I (want ...

2017-01-31T23:19:13.088+00:00

Wow! That was fast!
Some feedback on how I (want to) use it (I'm doing this on paper currently):

- I make daily predictions of tasks I'm planing for the day. In most cases the prediction is for the next day and thus the due time is always the next morning. I guess many people might do the majority of their predictions with a fixed schedule. Maybe support a default day/time?

- Many predictions I do I copy from the previous day - namely if I didn't complete them. I make a new prediction whether I will make it this day (often with a different confidence). I guess many people will make comparable predictions for different days/weeks/months. Maybe an option to copy a prediction (with auto-forwarded due-date) would be nice.

Thanks for reporting the issue. I've added a f...

2017-01-31T12:16:43.556+00:00

Thanks for reporting the issue. I've added a fix now, you'll probably get an update from Google Play within the next day or so.

Great approach. I installed the app (great that it...

2017-01-31T10:28:18.390+00:00

Great approach. I installed the app (great that it is so compact) and starting to experiment with it. One thing is very disappointing: If I create a new entry and forget to set a date or time I get a warning and the prediction is lost :-( Also I got a message that the time was too early though it wasn't and I also lost the prediction.