Not sure what algorithm is used and what data mining is done to give prediction analysis. Hope CRIS has considered journey starting and journey ending station as well in the algorithm, apart from only WL type. Because probability depends on stations as well, not just number. Even within same WL, a short journey with higher WL can get confirmed, while a long journey with lower WL may not get confirmed. In the below example, my CSMT-PUNE WL30 ticket can get confirmed, while my friend's CSMT-SUR WL20 ticket may remain waitlisted post charting.
--
I...
more... did some sample testing. Train 11027, date 03-Jun-2018, from CSMT to various stations like PUNE, DD, SUR, GR, WADI. Waiting List (PQ) and number (57/42) is same for availability till all these stations for SLEEPER.
--
For CSMT-PUNE, it shows probability 72%. However on most dates, all WL upto Pune get confirmed and few berths remain vacant after charting. Then how can this be 72%. It should be atleast 95%.
--
For CSMT-DD, it shows probability same as PUNE - 72%. Definitely DD journey has lower probability of getting confirmed than a PUNE journey in same WL. It appears that as per model, prediction is station independent, while in reality, it all depends on source-journey stations.
--
For CSMT-GR, it shows probability 73%. How can GR have a higher probability than PUNE for the same WL number?
--
For this particular scenario, if users rely on the prediction mode, they may get scared and not even book ticket till Pune and may actually miss out on journey, whereas the ticket would have got confirmed had it been booked.