Author: vilenkin

Normalized number of Urla duplicates on all hosts

CtrVideo(1 hour) / (CtrVideo(1 hour) + CtrVideo(8 hours))

Number of corrected clicks on documents from VIDEO divided by number of corrected shows for documents from VIDEO for given query during last 8 hours

Number of clicks on documents from VIDEO divided by number of shown documents from VIDEO for given query during last hour

Number of clicks on documents from VIDEO divided by number of shown documents from VIDEO for given query during last 8 hours

Number of clicks on documents from VIDEOQUICK divided by total number of clicks for given query during last 8 hours

For each SERP with given query in last 8 hours, calculate CTR for VIDEOQUICK and accumulate. Same for VIDEO source. Ratio of this values is the value for this factor

Number of corrected clicks on documents from VIDEOQUICK divided by total number of corrected clicks for given query during last 8 hours

Number of clicks on documents from VIDEOQUICK divided by number of shown documents from VIDEOQUICK for given query during last 8 hours

Mean view time of docs from VIDEO source (last 1 hour)

Mean view time of docs from VIDEOQUICK source (last 8 hours)

Short views proportion of docs from VIDEO source (1 hour / (1 hour + 8 hours))

Clicks on docs from VIDEO source ratio (1 hour / 8 hours)

CorrCtr is corrected clicks divided by corrected shows. Factor is CorrCtr(1 hour) / (CorrCtr(1 hour) + CorrCtr(8 hours)) of VIDEO source

Views of docs from VIDEO source (1 hour)

Shows of docs from VIDEOQUICK source (8 hours)

Shows(1 hour) / Shows(1 hour) + Shows(8 hours)) for VIDEOQUICK source

VIDEOQUICK share in total view time of service (last 8 hours)

VIDEOQUICK share in total log view time of service (last 8 hours)

Number of such requests (with same vnorm_dopp) during 1 hour divided by number of requests during 8 hours

Surplus on following data: block of fresh videos is a win block, each non-fresh video is loss block. One win (or loss) is a click on video from a block. Factor is a sum for each query sum for each block of surplus value for last hour

Surplus on following data: block of fresh videos is a win block, each non-fresh video is loss block. One win (or loss) is a click on video from a block. Factor is a sum for each query sum for each block of surplus value for last 8 hours

Surplus on following data: block of fresh videos is a win block, each non-fresh video is loss block. One win (or loss) is a click on video from a block. Factor is a sum for each query mean for all blocks of surplus value for last 8 hours

Surplus on following data: block of fresh videos is a win block, each non-fresh video is loss block. Win (or loss) is a view depth of videos from a block. Factor is a sum for each query sum for all blocks of surplus value for last 8 hours

Ratio between clicks on fresh results in last half hour and hour

Ratio between fresh surplus and total number of queries

Ratio between wins and wins + losses for fresh results

Ratio between clicks on news results in last half hour and hour

Ratio between news surplus and total number of queries

Ratio between wins and wins + losses for news results

Same as QuickClicksLastHourRatio, calculated using doppelgangers

Same as QuickSurplusToQueriesRatio, calculated using doppelgangers

Same as QuickWinsToTotalRatio, calculated using doppelgangers

Same as NewsClicksLastHourRatio, calculated using doppelgangers

Same as NewsSurplusToQueriesRatio, calculated using doppelgangers

Same as NewsWinsToTotalRatio, calculated using doppelgangers

Decayed counter for this query for last day, using doppelgangers instead of exact query

Same as QuickClicksLastHourRatioDopp, collect logs for 5 minutes before each aggregation

Same as QuickSurplusToQueriesRatioDopp, collect logs for 5 minutes before each aggregation

Same as QuickWinsToTotalRatioDopp, collect logs for 5 minutes before each aggregation

Same as NewsClicksLastHourRatioDopp, collect logs for 5 minutes before each aggregation

Same as NewsSurplusToQueriesRatioDopp, collect logs for 5 minutes before each aggregation

Same as NewsWinsToTotalRatioDopp, collect logs for 5 minutes before each aggregation

Ratio between clicks on images results in last half hour and hour, collect logs for 5 minutes before each aggregation

Ratio between images surplus and total number of queries, collect logs for 5 minutes before each aggregation

Same as QuickWinsToTotalRatio, collect logs for 5 minutes before each aggregation

Ratio between clicks on video results in last half hour and hour, collect logs for 5 minutes before each aggregation

Same as NewsSurplusToQueriesRatio, collect logs for 5 minutes before each aggregation

Same as NewsWinsToTotalRatio, collect logs for 5 minutes before each aggregation

Same as QuickClicksLastHourRatioDoppFixed, but use data only from mobile search

Same as QuickSurplusToQueriesRatioDoppFixed, but use data only from mobile search

Same as QuickWinsToTotalRatioDoppFixed, but use data only from mobile search

Aggregated sum of subqueries' fresh intent

Aggregated sum of subqueries' QDF (quick doc fraction)

Average fresh intent of subqueries

Average QDF of subqueries

Aggregated sum of subqueries' surplus (desktop)

Aggregated sum of subqueries' surplus (mobile)

Average surplus of desktop subqueries

Average surplus of mobile subqueries

Percent of requests, containing good hosts for this query during last 3 days

Replacement for FI_PAIRS_CONTRAST (desktop version)

Replacement for FI_PAIRS_CONTRAST (mobile version)

Replacement for FI_RT_CONTRAST_QUERY (desktop version)

Replacement for FI_RT_CONTRAST_QUERY (mobile version)

Ratio between clicks on fresh results in last half hour and hour, like QuickWinsToTotalRatio from blender (mobile version)

Ratio between fresh surplus and total number of queries, like QuickSurplusToQueriesRatio from blender (mobile version)

Ratio between wins and wins + losses for fresh results, like QuickWinsToTotalRatio from blender (mobile version)

Ratio between clicks on fresh results in last half hour and hour, calculated using doppelgangers, collect logs for 5 minutes before each aggregation (mobile version)

Ratio between fresh surplus and total number of queries, calculated using doppelgangers, collect logs for 5 minutes before each aggregation (mobile version)

Ratio between wins and wins + losses for fresh results, calculated using doppelgangers, collect logs for 5 minutes before each aggregation (mobile version)