Research assistant: build a small dataset on NBA coaching careers

I run a research project on how coaching careers advance in the NBA. Our dataset of coaching staffs starts with the 2006 season. For 169 coaches, 2006 is the first season we can see them, but most of them were already working in the NBA before that. I need someone to reconstruct the complete record of every NBA job those 169 people held before 2006. You will get an Excel workbook with a block of rows for each coach, already labelled with their name and their 2006 position. Many of those rows arrive already filled in; more on that below. One row goes in for each distinct job they held: first season, last season, team, the title as the source words it, a numeric rank where the job is a coaching job, what kind of job it was, a confidence flag, a source URL, and notes. I want every NBA job, not only the coaching ones. Video coordinator, scout, player development, and front-office roles all count and all get recorded as what they were. That distinction is the point of the exercise, so the job-kind column matters more than any other single field. 169 coaches. 846 job rows are already in the workbook as a starting point and I expect the finished file to hold somewhere between 1,100 and 1,400. Some coaches will take five minutes, some will take half an hour, and a few jobs will not be findable at all, which is a legitimate answer as long as you record it. ## What is already filled in, and what it is not Before posting I ran every one of the 169 names against Wikipedia and pulled the coaching career out of the infobox where there was one. That produced 846 rows, shaded pale blue in the workbook, covering 161 of the coaches. The other 8 arrive blank, either because no article exists or because the article is a playing biography that records no coaching career. Read the next paragraph carefully, because it is the part people get wrong. **A blue row is a lead, not an answer, and the blue rows are not a complete list.** Wikipedia infoboxes are reasonably good on head-coaching stints and quite bad on assistant and staff jobs, because an infobox line exists when somebody thought the job notable. The jobs this project is about are mostly the ones nobody thought notable. So on a Tier A coach you are doing three separate pieces of work: confirming the blue rows against a real source, correcting or deleting the ones that are wrong, and searching for the jobs Wikipedia left out. The third piece is the one that matters most to me. A coach who comes back with their five blue rows neatly sourced and three missing assistant jobs is not finished, and I will send them back. Most of this job is that third pass. About 61 of the 161 seeded coaches resolved to a Wikipedia title other than their plain name - Sam Mitchell (basketball), Mike Dunleavy Sr., and so on - because common names are common. Those rows carry a note reading CONFIRM THIS IS THE RIGHT PERSON in the tier list I will send you. Check the birth year or the team history on each before you trust the rows beneath it. A coach whose seeded rows belong to a different man of the same name is the worst outcome this job can produce, and it is the one I cannot spot afterwards. Two columns are deliberately blank on every blue row: rank, and kind of job. No script should guess those and mine did not try. They are yours to set from the source. ## Standards **Sources.** Every finished row needs a URL that actually shows the position and the season. A row without a source is not finished. Work in two passes, because the two halves of this job are not equally hard. The first pass is head coaching jobs. Basketball-Reference keeps a complete register of NBA head coaches with the seasons and teams, and it is trustworthy. Sweep every coach against it, record the head coaching jobs, and you will have perhaps a fifth of the rows in a fraction of the total time. The second pass is everything else, and it is where the work actually lives. Basketball-Reference does not carry assistant and staff records; its own coach pages carry the line "Assistant Coach and Staff records might not be complete." Your main instrument for the second pass is **prosportstransactions.com**, which does log assistant-coach hirings, dated to the day. I have confirmed it carries entries of the form "hired as assistant coach" going back into the 1990s. Supplement it with newspaper archives and contemporaneous coverage of staff hirings. Pre-2006 team media guides exist but are held in physical archives, so do not plan on reaching them. **Read the identity tags on prosportstransactions carefully, because they are not always right.** The site distinguishes same-named people by birth date or a parenthetical tag. On the test case I ran, a 1997 assistant-coach hiring was tagged with the birth date of a different, same-named man who was an active player at the time and who signed a playing contract with another team four months later. One of those two facts has to be wrong. Where a transaction's identity tag contradicts the surrounding record, say so in the notes and mark the row Uncertain rather than picking one. I would rather see the contradiction than a resolution you guessed at. Names repeat, and this is the single most likely way a row goes wrong. There is more than one Mike Brown in NBA records, more than one Mike Dunleavy, more than one Larry Brown. Check the birth year or the team history before trusting any page or transaction row. Two people with one name merged into one career is the error I am least able to detect downstream and most need you to prevent. **Not guessing.** Where the record is ambiguous, mark the row Uncertain and write what you found in the notes. I would rather have 800 clean rows and 150 flagged than 950 rows where I cannot tell which ones to trust. Guessed rows cost me more time than blank ones. **Reading the instructions carefully.** Several distinctions decide whether a row is right: NBA jobs count and college, G League, and overseas jobs do not; coaching jobs and staff jobs both count but are recorded differently; playing careers are not jobs for our purposes. The workbook explains each one and works through a full example. ## The person I need - Careful, patient work with records. This is not a volume task. - Comfort with Excel, including dropdown columns and not overwriting locked columns. - Enough familiarity with basketball to recognize a coaching title, or the willingness to learn it in the first hour. You do not need to be a fan. - Written English good enough to explain an ambiguous case in two sentences. ## Working process 1. I send the workbook and the instructions. 2. You complete the first 10 coaches and send the workbook back before going further. I check them and we fix any misunderstanding while it is cheap. 3. You complete the rest, sending progress in batches of roughly 25 coaches. 4. I review, come back with questions on the flagged rows, and we close out. Send questions in batches rather than one at a time. If the same question applies to many coaches, tell me. That usually means my instructions need fixing rather than that you did something wrong. ## To apply Do not send a generic cover letter. Instead, answer this in your proposal: Mike Brown was the head coach of the Cleveland Cavaliers in 2006. What was his first NBA coaching job, which season was it, and what is your source? It should take about five minutes. I read the answers, and an application without one goes no further. Also tell me, in one line each: how many hours a week you can give this, and whether you have done records or archival research before. ## Terms Two stages, because I would rather measure this job than guess at it. The 169 coaches split into two tiers, and you will be told which tier each is in before you start. 161 are Tier A: a usable Wikipedia article exists, so the work is confirming the seeded rows, correcting them, and finding what the article left out. 8 are Tier B: no usable article, so the whole career is built from team media guides and newspaper archives. Stage 1, the pilot. An hourly contract at $22 per hour capped at 4 hours, covering 10 coaches, 9 Tier A and 1 Tier B. I am paying for a measurement here as much as for the rows: at the end I will know the real minutes per coach in each tier, and so will you. Stage 2, the remaining 159 coaches, 152 Tier A and 7 Tier B. A fixed-price contract at a per-coach rate in each tier, set from what the pilot actually showed. My working assumption going in is about 12 minutes for a Tier A coach and about 25 for a Tier B coach, which is what the rates below are built on and works out to roughly $22 an hour either way. The Tier B figure allows for confirming identity on every row, which the sources make necessary. If the pilot says otherwise the rates move, in your favour as readily as mine. - Tier A: $4.50 per coach completed. 152 coaches, $684.00. - Tier B: $9.00 per coach completed. 7 coaches, $63.00. - Completion bonus: 10% of the stage 2 total, paid at the end if the finished workbook holds up on review. Flagged rows do not count against it. Guessed rows do. Stop at 30 minutes. If a Tier B coach has taken half an hour and you still cannot place them, mark the row Cannot find, write down where you looked, and move on. You are paid the same for that coach as for an easy one. I would rather buy 30 minutes of honest searching than two hours of it, and a handful of these people left almost no record. A coach counts as completed when every job they held before 2006 has a row, every row has a source, every blue row has been confirmed or corrected, and anything ambiguous is flagged. Work is remote and on your own schedule. The data is public record; nothing here is confidential.