Question: Wondering if we have historical success data on Experimental mode

[SamuelDovgin]

Hi! When I use the application I'm usually confused if its best to turn the experimental model on or off as the resulting TER can drastically differ. I was spending some time looking into the "Multinomial Logit (Logistic) Model" that it uses to predict probabilities however I didn't get that far as the math seems quite complex.

However I was wondering if we have data to suggest that the "Experimental model" has a better prediction rate and overall outcome than just leaving it off? or is it fully experimental and cannot be relied upon (and if its better to use the legacy model if we just want consistent TER)

thanks!