We're partnered with many of the federations around the world and from time to time will perform a full, historic import of their results as each new federation comes on board. This is quite a complex task with the goals being:
- Import all their results.
- Calibrate their players over space and time so that they all have a full, accurate, consolidated level history that is comparable to everyone else on the system.
- *Not* cause calibration shock to those already on the system. I.e Although we recognise some players will be affected, we aim to minimise the overall effect of importing hundreds of thousands - sometime millions - of historic results on everyone else.
To acheive this, we do the following:
- Work with the federation to access historic results via a CSV file (or files) that is as close to our standard format as we can go.
- Import all the results into our developement system which is a copy of the live system at the time. We generally don't import large numbers of new results straight into our live system.
- Do an initial pass looking for duplicate players - i.e. players that have a results profile from the new federation and one from the existing system. These are usually the higher level players becauce they are more recognisable and they are also the ones that travel most. By merging these duplicates, we give the system it's first steer on player level, albeit from the top down.
- We then let the calibration work it's magic which can take a while as it only makes small adjustments across the system to minimise the overall calibration error.
- Once the calibration dust starts to settle we then run our duplicate detection tool which actively looks for players at all levels who have profiles for the new results also also existing profiles . We're looking for anythin that suggests they are the same player and it presents a list of possiblities that we review manually and merge if we genuinely believe they are duplicates. This can be quite time consuming as we're often presented with thousands of potential duplicates to check!
- With most of the duplicates found and merged we let the calibration run again and settle.
- At this point we review our rankings with the federation. They're usually pretty close to expectation by now but there can be some surprises and we chase those down to understand what the issues might be. They could be:
- Duff results - especially early duff results which can have a significant effect on overall levels
- Incorrectly merged players
- Actually correct but that players are changing so fast current thinking was behind current level
- Whole clubs or regions need adjusting up or down to match the rest of the world.
- We also check to see if existing players have been affected by this import and calibration. It's likely that some players will be affected to some extent but we try to minimise that.
- We do have some tools to help us analyse and resolve these cases. We're usually looking at just a few players by this time though it could be a whole region.
- When everyone is happy, we export the results, along with all the information we've recorded about duplicates and calibration over to the live system. By migrating all this additional calibration data, we hope to add all the results from the new federation with minimal impact on the existing players on the system.
As you can see this is quite an intensive process and we try to minimse the impact on those already on the system. There will still be some players affected though. Those who play against the newly imported players such as masters can be more affected than most.
We will warn you that we are onboarding a new federation so please be ready for some amount of change. Obviously, if you feel hard done by, do get in touch and we'll investigate!