Winner's Curse - Revision history

JakeSilberman: JakeSilberman moved page Winner's Curse to Governance Measures

2016-07-27T17:10:29Z

JakeSilberman moved page Winner's Curse to Governance Measures

Show changes

imported>Ed: /* Data */

2013-04-23T22:11:46Z

Data

imported>Ed at 01:37, 24 January 2013

2013-01-24T01:37:19Z

imported>Ed at 01:36, 24 January 2013

2013-01-24T01:36:13Z

imported>Ed: /* Skaters */

2012-08-20T18:14:03Z

Skaters

imported>Ed: /* Contract Data */

2012-08-20T05:45:45Z

Contract Data

imported>Ed: /* Contract Data */

2012-08-20T05:40:15Z

Contract Data

imported>Ed: /* Contract Data */

2012-08-20T04:59:27Z

Contract Data

imported>Ed: /* NHL Data */

2012-08-20T04:49:38Z

NHL Data

imported>Ed: /* CapGeek Data */

2012-08-20T04:47:30Z

CapGeek Data

← Older revision		Revision as of 22:11, 23 April 2013
Line 34:		Line 34:

	Of the 2865 NHL players with statistics, we have CapGeek contract records for 1455 (51%). Part of the discrepancy is due to retirement. 442 players played their last game before 2001 and do not have CapGeek records. This leaves 968 (34%) of the NHL players that do not have contracts in CapGeek for some other reason - most likely that CapGeek's coverage is incomplete.		Of the 2865 NHL players with statistics, we have CapGeek contract records for 1455 (51%). Part of the discrepancy is due to retirement. 442 players played their last game before 2001 and do not have CapGeek records. This leaves 968 (34%) of the NHL players that do not have contracts in CapGeek for some other reason - most likely that CapGeek's coverage is incomplete.
		+
		+	====Team Level Data====
		+
		+	The crawler was modified to collect team level data, from pages like:
		+	http://www.nhl.com/ice/teamstats.htm?season=20122013&gameType=2&viewName=summary
		+
		+	Data was collected for all teams for all seasons (1997-1998 to 2011-2012, except 2004-2005), for both the regular season and the playoffs (for teams that made the playoffs). The following variables were joined to the dataset at the (team1,season) level:
		+
		+	*teamcode -- three letter team code for the join
		+	*teamseason -- season xxxxyyyy for the join
		+	*teamgp -- games played (reg season)
		+	*teamp -- points (reg season)
		+	*teamgperg -- goals per game (reg season)
		+	*teamgaperg -- goals against per game (reg season)
		+	*po_teamgp -- games played (playoffs)
		+	*po_teamp -- points (playoffs)
		+	*po_teamgperg -- goals per game (playoffs)
		+	*po_teamgaperg -- goals against per game (playoffs)
		+
		+	Note that some players are recorded for multiple teams during a single season. This recording is reflected in the team1, team2, team3 and team4 variables (no player played for more than four teams), as well as in the teamstring variable (which lists the teams as a comma separated string) and the teamcount variable (which counts the number of teams that a player played for). The join was done at the team1 level. As such, the team data will be 'incorrect' (or at least potentially unfairly allocated) for players that have a teamcount > 1.

	==Variables==		==Variables==

@@ Line 7: / Line 7: @@
 The Postgres database hosting the data is on my Sauder workstation (137.82.145.70) and is named 'NHL' (note the caps). You can log in remotely (providing my VPN is down) from powershell or any psql capable client with:
-  psql -h 137.83.145.70 -u Ed NHL
+  psql -h 137.83.145.70 -U Ed NHL
   bottle

← Older revision		Revision as of 01:36, 24 January 2013
Line 5:		Line 5:

	==Data==		==Data==
		+
		+	The Postgres database hosting the data is on my Sauder workstation (137.82.145.70) and is named 'NHL' (note the caps). You can log in remotely (providing my VPN is down) from powershell or any psql capable client with:
		+	psql -h 137.83.145.70 -u Ed NHL
		+	bottle

	The two primary datasets are:		The two primary datasets are:

@@ Line 251: / Line 251: @@
 Penalties (pen)
   Minor - Minor Penalties
   Major - Major Penalties

@@ Line 101: / Line 101: @@
 *cyear - contract year signed (i.e., first contract year)
 *cage - contract age for a given year (i.e. 1 for first year, 2 for second year, etc).
 *cgteam - the team code of the team that holds the player's contract (according to CapGeek)
 *playedcteam - 1 if the player played for the contract holder in the season, 0 otherwise

@@ Line 114: / Line 114: @@
 *relcontractid - unique relative contract ids (i.e., within player sequential ids).
 *firstcontract - 1 if this is the first contract listed, 0 otherwise
-*typechanged - 1 if the type of this contract is different from the type of the previous contract
+*typechanged - 1 if the status (RFA/UFA/etc) of this contract is different from the type of the previous contract
 *rfa2ufa - 1 if the transistion was from RFA last contract to UFA this period, 0 otherwise
 *ent2rfa
@@ Line 120: / Line 120: @@
 *unk2rfa
 *unk2ufa
-*transitioncode - 0=No Transition, 1=ENT->RFA, 2=ENT->UFA, 3=RFA->UFA, 4=UNK->RFA, 5=UNK->UFA, 9=UFA->RFA
+*transitioncode - 0=No Transition, 1=ENT->RFA, 2=ENT->UFA, 3=RFA->UFA, 4=UNK->RFA, 5=UNK->UFA, 9=UFA->RFA, -1=The previous contract status can't be deduced, because the current contract is the first listed.
 Note that if a player played for the Atlanta Thrashers and the Winnipeg Jets, this counts as the same team.
 ===Player's Data===

@@ Line 23: / Line 23: @@
 ===NHL Data===
-The NHL data was retrieved from NHL.com using a custom webcrawler. The data covers the years 1997-present, except for the 2004-2005 season (which was a lockout). Data was retrieved seperately for skaters and goalies. We have data on all 2870 NHL players that have stats recorded on NHL.com in the final dataset.
+The NHL data was retrieved from NHL.com using a custom webcrawler. The data covers the years 1997-present, except for the 2004-2005 season (which was a lockout). Data was retrieved seperately for skaters and goalies. We have data on all 2865 NHL players that have stats recorded on NHL.com in the final dataset.
 ====Joining to CapGeek====
@@ Line 29: / Line 29: @@
 NHL.com does not provide unique player identifiers. These were created using player names and dates of birth. Players were matched to CapGeek using their names, and when names were non-unique matches were established manually by checking birth dates, teams and positions in both sources. Birth-date - player-name dyads were found to be unique. For a small percentage of players (approximately 45/1459) their names were recorded differently in each source. Fuzzy-matching software provided potential candidate matches and all matches were checked by hand.
-Of the 2870 NHL players with statistics, we have CapGeek contract records for 1459 (51%). Part of the discrepancy is due to retirement. 442 players played their last game before 2001 and do not have CapGeek records. This leaves 969 (34%) of the NHL players that do not have contracts in CapGeek for some other reason - most likely that CapGeek's coverage is incomplete.
+Of the 2865 NHL players with statistics, we have CapGeek contract records for 1455 (51%). Part of the discrepancy is due to retirement. 442 players played their last game before 2001 and do not have CapGeek records. This leaves 968 (34%) of the NHL players that do not have contracts in CapGeek for some other reason - most likely that CapGeek's coverage is incomplete.
 ==Variables==

@@ Line 19: / Line 19: @@
 *The full contract history. Notes on contracts (i.e. known clauses) are included in the dbase but not the STATA export.
-CapGeek provides contract data on 2241 players, of which 1459 appear in the NHL statistics data, who held contracts from 2001 to the present day. 194 capgeek players have teams marked as 'Inactive', and these invariably do not appear in the NHL data. The missing capgeek players presumably play in the minors.
+CapGeek provides contract data on 2241 players, of which 1455 appear in the NHL statistics data, who held contracts from 2001 to the present day. 194 capgeek players have teams marked as 'Inactive', and these invariably do not appear in the NHL data. The missing capgeek players presumably play in the minors.
 ===NHL Data===