We cannot yet set a scan against the AHledger daily figure, and one scan alone is a noisy reading
Este conteúdo não está disponível em sua língua ainda.
We asked how far one of our own scans sits from the AHledger daily figure for the same day. It matters because the lists of daily movers lean on the newest scan. If part of a one day move is only the way the day was measured, a guild member should not act on it.
The question
Section titled “The question”Our price record is made of two kinds of day. A daily figure is one price per item for a whole day, taken from AHledger’s daily history. A scan is one reading of the full price table at one moment. For days with scans, the last scan of the day stands for that day.
The two are not measured the same way. So when an item “moves” from one day to the next, some of that move may be measurement and not the market. A useful answer would be a plain figure: on the same day, for the same item, a scan sits this far from the daily figure. Failing that, a useful answer says how much trust a single scan has earned.
The short answer
Section titled “The short answer”- We cannot measure the gap head on yet. The Horde record holds 16 days of AHledger daily figures and 27 days of our own scans, and no day has both kinds.
- The one place the two kinds meet is 8 October, our first scan, against the daily figure for 7 October. Of 303 Horde items priced on both days, 118 rose by a tenth or more and 128 fell by a tenth or more. One day earlier, between two daily figures, it was 119 up and 96 down out of 320. The join is a little rougher than an ordinary day, not wildly so.
- Big one day moves are normal whichever way the day is measured. Between our scans of 15 November and 16 November, 119 of the same 319 items rose by a tenth or more and 103 fell. A mover seen in a single scan deserves little weight on its own.
Method
Section titled “Method”We looked at the items AHledger tracks in its daily history: 323 items on PvP Horde and 325 on PvP Alliance. These are the only items that have both kinds of day in their record, so they are the only fair ground for a comparison.
The direct test needs a day with both a daily figure and a scan. Our record keeps the daily figure only for days with no scan, so there is no such day. Daily figures run from 21 September to 7 October. Scans run from 8 October onward. That leaves three indirect tests. Everything below this point is our own calculation.
The join. We compared 7 October (daily figure) with 8 October (scan), and set that next to ordinary pairs of days on either side. We counted an item as a mover when its middle price changed by a tenth or more. We chose a tenth because a smaller change is not worth acting on once the auction house cut is paid.
A scan against its own day. From 5 November we have many scans a day, about one an hour, 200 scans over 13 days on Horde. For each item and day we took the middle price across all of that day’s scans, then asked how far the last scan of the day sat from it. We only used item days with six or more priced scans. This middle of the day is built from our scans. It stands in for the AHledger daily figure. It is not the same thing.
Day middle against last scan. For the same 13 days we measured the typical move from one day to the next twice: once using the last scan of each day, once using the middle of each day.
What came out, on Horde:
- At the join, the typical item moved 28.1% from the day before. Between two daily figures the day before, it was 20.8%. Across all the AHledger days the typical move ran from 15.1% to 38.8%, so the join sits inside the ordinary range.
- The last scan of a day sat 6.5% from the middle of its own day for the typical item. About 41% of readings sat a tenth or more away, and about 20% sat a quarter or more away. One third sat within 2%.
- The typical one day move was 21.2% from last scan to last scan, and 15.7% from day middle to day middle. Alliance gave 21.9% and 15.9%.
So roughly a quarter of the typical one day move goes away when a day is summed up by its middle and not by its last scan. That quarter is our best estimate of the part that is “just measurement”. The rest is real churn in a small market.
Charts
Section titled “Charts”Each point is one day. Its height is how far the typical tracked item moved from the day before, up or down. The blue line is AHledger’s daily history. The first orange point is the join, the one day that compares a daily figure with a scan. It is high, but no higher than two of the blue days. The lines never settle near zero under either method.
Each bar is a share of readings. The top bars are scans that agree closely with their own day. The bottom three bars are scans that sit a tenth or more away. Add them up and about four readings in ten are that far off, on a day where nothing but the hour changed.
The longer bar in each pair uses the last scan of each day. The shorter bar uses the middle of each day. The gap between them is the part of a one day move that comes from picking one moment to stand for the day. Both factions tell the same story.
What would change this
Section titled “What would change this”- A real overlap. If AHledger’s daily figures were kept for days that also have our scans, for a week or more, the question could be answered head on. If the typical scan then sat within a few percent of the daily figure, single scans would deserve more trust than we give them here. If it sat a tenth or more away, the join on 8 October would look like a measurement jump after all.
- Calmer days. If the typical one day move between day middles fell well under a tenth, for example as the market grows after launch, then a single scan showing a move of a quarter would stand out and mean something.
- A different stand in. We used the last scan of the day because that is what the record uses. If a later check shows the last scan is taken at an odd hour, such as the quiet early morning, the gap could be larger than our estimate.
Until then, treat a mover on Top 10 today as a prompt to look, not a fact. A move that shows up across several days on Top 10 this week or on the Charts page has earned more weight. The Methodology page explains how the prices are built.
Limits of beta data
Section titled “Limits of beta data”- No overlap. The core question is not answered. Everything here is indirect: one join day, and our own scans compared with themselves.
- One join day. The comparison of a daily figure with a scan rests on a single pair of days, 7 and 8 October. One day can be odd for reasons of its own.
- Thin history. Many scans a day exist for only 13 days. No scans ran from 22 October to 4 November. October scan days have one scan each, so they cannot show how a scan sits inside its own day.
- Small population. Few sellers means one person posting or cancelling can shift a middle price. That is real churn, but it is beta churn.
- Level 30 cap and no raids. The items traded, and how often, will be different at launch.
- The wipe. The beta is wiped. None of these percentages carry over. What may carry over is the habit: one reading of a small market is a weak signal.
- Tracked items only. The comparison covers the roughly 320 items in AHledger’s daily history for each faction. These are the busier items. The thousands of rarely listed items in a full scan may behave differently.
- Estimates. Sold a day is an estimate and exists only in the newest scan, so we could not sort items by how fast they sell on past days.