Phidias1
異星工廠（Factorio）的每秒更新次數（UPS）效能對於大型基地設計至關重要。本論文比較了每台組裝機周圍 4 組件廣播塔與 8 組件廣播塔兩種配置在兩種基地下的 UPS 影響： 等量科研產出與 等量實體數量。測試使用三張地圖：地圖 A（4 組件廣播塔，基線）、地圖 B（8 組件廣播塔，與 A 科研產出相同）和地圖 C（8 組件廣播塔，與 A 機器數量相同）。利用異星工廠無頭伺服器的基準模式，記錄了每張地圖執行 12,000 tick 的模擬時間，每個條件進行 30 次獨立執行。透過四分位距（IQR）法移除異常值，採用 Kruskal‑Wallis 檢定（α = 0.01）及 Dunn 事後兩兩比較。在等量科研產出下，8 組件廣播塔設計比 4 組件廣播塔設計 執行時間縮短 41.0%（p < 0.001，d = 33.66）；在等量機器數量下，8 組件廣播塔設計比 4 組件廣播塔設計 執行時間縮短 5.5%（p < 0.001，d = 4.87）。
Factorio’s updates per second (UPS) performance is critical for large scale base design. This thesis compared the UPS impact of 4 versus 8 beacons per machine under two bases: equal science output and equal entity count. Three maps were tested: Map A (4 beacon, baseline), Map B (8 beacon, same science as A), and Map C (8 beacon, same number of machines as A). Using Factorio’s headless server in benchmark mode, we recorded simulation time for 12,000 ticks across 30 independent runs per condition. Outliers were removed via the interquartile range method. A Kruskal‑Wallis test (α = 0.01) with Dunn’s post hoc comparisons was applied. For equal science output, the 8 beacon design was 41.0% faster in execution time improvement than 4 beacon (p < 0.001, d = 33.66). For equal machine count, the 8 beacon design was 5.5% faster in execution time (p < 0.001, d = 4.87).
異星工廠是一款工廠自動化遊戲，玩家建造的工廠可擴展至極其龐大的規模。隨著工廠的擴張，模擬引擎的計算需求也隨之增加，常常導致遊戲的每秒更新次數（UPS）低於標稱的 60 UPS。此時，遊戲的執行速度慢於即時，成為進一步擴張的瓶頸。因此，最佳化 UPS 已成為異星工廠社群的核心關注點，尤其是在生產數千瓶每分鐘科研包的超大型基地中。
Factorio is a factory automation game where player built factories can grow to enormous scales. As the factory expands, the computational demand on the simulation engine increases, often causing the game’s updates per second (UPS) to drop below the nominal 60 UPS. At that point, the game runs slower than real time, creating a bottleneck for further growth. Consequently, optimizing UPS has become a central concern for the Factorio community, particularly in megabase play where players build factories producing thousands of science packs per minute.
一個常見的設計選擇是圍繞組裝機的組件廣播塔數量。組件廣播塔將模組效果傳遞給相鄰的機器，使得生產力或速度模組得以共享。兩種熱門配置是每台組裝機 4 組件廣播塔與每台組裝機 8 組件廣播塔。直觀上，8 組件廣播塔應提高單位面積產出，但其對 UPS 的影響（由於實體數量增加、機械臂活動增多以及輸送帶處理量上升）則不那麼明顯。
One common design choice is the number of beacons placed around assemblers. Beacons transmit module effects to adjacent machines, allowing productivity or speed modules to be shared. Two popular configurations are 4 beacons per machine and 8 beacons per machine. While intuition suggests that 8 beacons should improve production per area, the impact on UPS, due to increased entity count, inserter activity, and belt handling, is less obvious.
我們比較了等量每秒科研產出（匹配總產量）和等量實體數量（匹配機器數量）。為此，準備了三個存檔檔案，使用空地圖。不涉及敵人、污染等其他因素。存檔間的唯一區別是組件廣播塔佈局：
We compare equal science output per second (matching total production) and equal entity count (matching the number of machines). To achieve this, three save files were prepared, using an empty map. No other factors such as enemies, pollution are involved. The only difference between the saves was the beacon layout:
核心假設是：在兩種比較下，8 組件廣播塔配置相較於 4 組件廣播塔配置都能帶來統計上顯著的 UPS 提升。我們採用 Kruskal‑Wallis 檢定 (Kruskal & Wallis, 1952) （非參數單因子變異數分析），α = 0.01，隨後進行 Dunn 事後檢定 (Dunn, 1964)，並報告 Cohen’s d 以量化實際顯著性 (Cohen, 1988)。完整的流程 (Peng, 2011)，包括異星工廠存檔檔案、基準測試腳本、分析程式碼及原始結果，均可在專案雲端儲存中取得。
The central hypothesis is that the 8 beacon configuration yields a statistically significant improvement in UPS compared to the 4 beacon configuration, under both comparisons. We adopt a Kruskal Wallis test (Kruskal & Wallis, 1952) (non‑parametric analogue of ANOVA) with α = 0.01, followed by Dunn’s post hoc tests (Dunn, 1964), and report Cohen’s d to quantify practical significance (Cohen, 1988). The complete pipeline (Peng, 2011), including Factorio save files, benchmarking scripts, analysis code, and raw results, is available on the project cloud.
所有基準測試在一台專用 Linux 服務器上執行，採用異星工廠2.0.76 版本無頭伺服器（headless server）的基準模式，遊戲僅進行模擬，無任何圖形或音訊輸出。
All benchmarks were executed on a dedicated Linux server. The benchmark mode is used in the Factorio headless server version 2.0.76, where the game simulates without any graphical or audio output.
| 硬體 Hardware |
|
| 系統 System |
Ubuntu 24.04.4 LTS |
| 主板 Motherboard |
MSI S3991 |
| 處理器 CPU |
AMD R7‑9700X 8 核心 Core 頻率 Frequency 3.8 (5.5) GHz |
| 記憶體 RAM |
MICRON UDIMM 2x32 GB 64 GB DDR5-5600 CL 46 |
| 固態硬碟 SSD |
Samsung PM9A1 NVMe |
每次測量包含以下步驟：首先啟動無頭伺服器並載入地圖；接著執行總計 12,000 tick（在 60 UPS 下相當於 200 秒）的模擬；最後停止伺服器，並重複上述步驟進行下一次執行。總共執行 30 次獨立執行，按固定順序（ABCABCABC…）進行。該樣本量滿足中央極限定理，並提供足夠的統計功效以檢測差異。
Each measurement consists of the following steps: first, start the headless server and load the map. Then, run the simulation for a total of 12,000 ticks (200 seconds at 60 UPS). Finally, stop the server and repeat the process for the next run. A total of 30 independent runs were performed, at a fixed order (ABCABCABC...). This sample size was chosen to satisfy the central limit theorem and provide adequate statistical power for detecting a difference.
資料收集後，我們對每個條件分別應用了四分位距（IQR）法：首先計算每個條件的 Q1 與 Q3，接著定義界限為 Q1 − 1.5 × IQR 與 Q3 + 1.5 × IQR，任何平均 UPS 落在此區間之外的執行則視為異常值並從分析中移除。最終，我們使用 IQR 法從地圖 C 中移除了一個異常值，最終樣本量為：A 組 n = 30，B 組 n = 30，C 組 n = 29。
After data collection, the Interquartile Range (IQR) method was applied to each condition: Q1 and Q3 were calculated, fences were defined as Q1 − 1.5 × IQR and Q3 + 1.5 × IQR, and any run falling outside these fences was removed as an outlier. This resulted in the removal of one outlier from Map C, yielding final sample sizes of n = 30 for Map A, n = 30 for Map B, and n = 29 for Map C.
所有分析使用 Python 3.12 完成。由於 Shapiro‑Wilk 檢定表明殘差非常態 (Shapiro & Wilk, 1965)，我們採用：
All analyses were performed in Python 3.12. Because the Shapiro‑Wilk test indicated non‑normal residuals (Shapiro & Wilk, 1965), we used:
由於資料不服從常態分佈，進行了 Kruskal‑Wallis 檢定 (Kruskal & Wallis, 1952)。檢定顯示三張地圖間存在高度顯著差異（H = 78.225，p < 0.001）。隨後進行了帶有 Bonferroni 校正的 Dunn 事後兩兩比較 (Bonferroni, 1936; Dunn, 1964)，結果列於表 1。所有差異在 α = 0.01 水準上均具有統計顯著性。
Because the data was not normally distributed, a Kruskal‑Wallis test was performed (Kruskal & Wallis, 1952). The test revealed a highly significant difference among the three maps (H = 78.225, p < 0.001). Dunn’s post hoc pairwise comparisons were conducted with Bonferroni correction (Bonferroni, 1936; Dunn, 1964). Results are presented in Table 1. All differences were statistically significant at α = 0.01.
效應量最大的比較是 A 對 B（等量科研產出）：8 組件廣播塔設計（B）比 4 組件廣播塔設計（A）快 41.0%。效應量 d = –33.66，極大，證實了實體數量減少本身就能帶來巨大的 UPS 提升。
比較 A 對 C（等量機器數量）則分離了組件廣播塔密度的影響。地圖 C（8 組件廣播塔）比地圖 A（4 組件廣播塔）快 5.5%，Cohen’s d = –4.87。該效應量按常規標準仍屬極大，但遠小於 A vs B 的差異，表明 UPS 提升的主要來源是機器數量的減少，而非組件廣播塔佈局本身。
比較 B 對 C（相同組件廣播塔類型，不同實體數量）進一步確認實體數量是主導因素：B（較少機器）比 C（較多機器）快 60.1%，d = 31.78。
The comparison with the largest practical effect was A vs B (equal science output), where the 8‑beacon design (B) outperformed the 4‑beacon design (A) by 41.0%. The effect size (d = −33.66) is extraordinarily large, confirming that the entity‑count reduction alone provides a massive UPS gain.
The comparison A vs C (equal machine count) isolates the effect of beacon density. Map C (8 beacon) was 5.5% faster than Map A (4 beacon), with a Cohen’s d of −4.87. While still a very large effect by conventional standards, it is substantially smaller than the A vs B difference, indicating that the bulk of the UPS improvement comes from reducing the number of machines, not from the beacon layout per se.
The comparison B vs C (same beacon type, different entity count) confirms that entity count is the dominant factor: B (fewer machines) was 60.1% faster than C (many machines), with d = 31.78.
| 地圖 Map |
差距 Difference |
標準差 Standard Deviation |
Cohen D | p |
| A (n = 30) vs B (n = 30) | -220.60 UPS (-41.000%) | 4.202 409 vs 8.260 598 | -33.662 | 0.000 000 |
| A (n = 30) vs C (n = 29) | -18.63 UPS (-5.543%) | 4.202 409 vs 3.399 014 | -4.865 | 0.000 035 |
| B (n = 30) vs C (n = 29) | 201.97 UPS (60.097%) | 8.260 598 vs 3.399 014 | 31.779 | 0.000 035 |
結果證實，在追求特定科研產出時，8 組件廣播塔設計在 UPS 方面具有強烈優勢。然而，新加入的地圖 C 對比揭示：即使機器數量相同，8 組件廣播塔佈局仍能帶來約 5.5% 的開銷降低，這源於不同的機器和組件廣播塔互動。很可能是因為更多組件廣播塔使每台機器更快達到生產力上限，從而減少了單位產出所需的機械臂擺動次數和輸送帶流量。
The results confirm that the 8‑beacon design is strongly preferred for UPS when aiming for a given science output. However, the new Map C comparison reveals that even when the number of machines is held equal, the 8‑beacon layout provides a modest (~5.5%) overhead reduction due to the different machine and beacon interactions. This is likely because with more beacons per machine, each machine reaches its productivity cap faster, reducing inserter swings and belt traffic per unit of output.
A vs C 的效應量 d ≈ 4.87 在實際中仍然足夠大，尤其對於超大型基地而言。地圖 B 的 UPS 變異性較大（SD = 8.26），表明 8 組件廣播塔佈局對遊戲內事件更加敏感。
The effect size of d ≈ 4.87 for A vs C is still large enough to be meaningful in practice, especially for megabases. The greater variability in Map B’s UPS (SD = 8.26) compared to A or C suggests that the 8‑beacon layout is more sensitive to in‑game events.
然而，本研究仍有其侷限性。首先，我們僅測試了一種基礎設計和一種模組配置（機器使用產能模組，組件廣播塔使用速度模組），因此結果可能無法直接推廣到其他模組組合或不同的基地佈局。其次，地圖 A 與地圖 C 之間 5.5% 的效能差異，在直插式（direct insertion）等高度最佳化的工廠設計中可能不復存在。最後，所有測試均在單一硬體上進行，雖然我們透過多次重複測量降低了隨機誤差，但仍無法完全排除硬體特定雜訊的影響。
However, this study has several limitations. First, we tested only one base design and one module configuration (productivity modules in machines with speed modules in beacons). Therefore, the results may not generalize to other module combinations or different base layouts. Second, the 5.5% performance gain observed between Map A and Map C may not hold in highly optimized factory designs such as direct insertion. Finally, all benchmarks were conducted on a single hardware platform. Although repeated measurements reduced random error, the influence of hardware-specific noise cannot be entirely ruled out.
總之，對於固定的科研目標，8 組件廣播塔設計明確更優；即使機器數量不變，它仍能帶來 5.5% 的增益。這些發現為組件廣播塔設計決策提供了統計上的堅實證據。未來工作應測試其他組件廣播塔數量、工廠規模以及物流選擇。
In conclusion, the 8‑beacon design is unequivocally better for fixed science goals and still yields a 5.5% gain even when machine count is unchanged. These findings provide statistically grounded evidence for beacon design decisions. Future work should test other beacon counts, factory scales, and logistics choices.
C. E. Bonferroni. (1936). Teoria statistica delle classi e calcolo delle probabilità. Pubblicazioni Del R Istituto Superiore Di Scienze Economiche e Commerciali Di Firenze, 8, 3–62.
Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd edn). Routledge. https://doi.org/10.4324/9780203771587
Dunn, O. J. (1964). Multiple Comparisons Using Rank Sums. Technometrics, 6(3), 241–252. https://doi.org/10.2307/1266041
Kruskal, W. H., & Wallis, W. A. (1952). Use of Ranks in One-Criterion Variance Analysis. Journal of the American Statistical Association, 47(260), 583–621. https://doi.org/10.2307/2280779
Peng, R. D. (2011). Reproducible Research in Computational Science. Science, 334(6060), 1226–1227. https://doi.org/10.1126/science.1213847
Shapiro, S. S., & Wilk, M. B. (1965). An Analysis of Variance Test for Normality (Complete Samples). Biometrika, 52(3/4), 591–611. https://doi.org/10.2307/2333709
| 地圖 Map |
回數 Run |
更新數 UPS |
運行時間 EXE Time |
開始時間 Start Time |
結束時間 End Time |
平均時間 Avg Time |
最小時間 Min Time |
最大時間 Max Time |
| ms | ||||||||
| A | 1 | 323.45 | 37,100.20 | 3.043 | 40.977 | 3.092 | 0.140 | 16.649 |
| B | 1 | 531.71 | 22,568.72 | 2.673 | 25.960 | 1.881 | 0.131 | 12.918 |
| C | 1 | 332.90 | 36,046.99 | 3.133 | 39.996 | 3.004 | 0.162 | 18.102 |
| A | 2 | 315.79 | 38,000.14 | 3.086 | 41.949 | 3.167 | 0.149 | 17.100 |
| B | 2 | 539.48 | 22,243.60 | 2.714 | 25.676 | 1.854 | 0.133 | 12.781 |
| C | 2 | 337.21 | 35,586.54 | 3.108 | 39.510 | 2.966 | 0.168 | 17.883 |
| A | 3 | 316.44 | 37,921.31 | 3.080 | 41.825 | 3.160 | 0.141 | 17.485 |
| B | 3 | 543.58 | 22,075.78 | 2.663 | 25.448 | 1.840 | 0.133 | 12.962 |
| C | 3 | 339.66 | 35,329.77 | 3.108 | 39.258 | 2.944 | 0.158 | 17.607 |
| A | 4 | 319.04 | 37,612.99 | 3.053 | 41.499 | 3.134 | 0.140 | 17.409 |
| B | 4 | 543.83 | 22,065.72 | 2.695 | 25.484 | 1.839 | 0.131 | 12.877 |
| C | 4 | 338.07 | 35,495.37 | 3.154 | 39.479 | 2.958 | 0.158 | 17.853 |
| A | 5 | 310.09 | 38,698.89 | 3.106 | 42.644 | 3.225 | 0.147 | 17.412 |
| B | 5 | 534.86 | 22,435.84 | 2.730 | 25.886 | 1.870 | 0.130 | 12.887 |
| C | 5 | 324.46 | 36,984.21 | 3.109 | 40.913 | 3.082 | 0.165 | 18.285 |
| A | 6 | 307.01 | 39,086.14 | 3.055 | 42.978 | 3.257 | 0.141 | 17.348 |
| B | 6 | 518.66 | 23,136.43 | 2.714 | 26.572 | 1.928 | 0.129 | 12.642 |
| C | 6 | 337.23 | 35,583.69 | 3.137 | 39.538 | 2.965 | 0.172 | 17.772 |
| A | 7 | 311.01 | 38,584.36 | 3.076 | 42.509 | 3.215 | 0.142 | 17.289 |
| B | 7 | 537.54 | 22,324.10 | 2.778 | 25.829 | 1.860 | 0.133 | 12.515 |
| C | 7 | 332.80 | 36,057.83 | 3.121 | 39.998 | 3.005 | 0.154 | 17.838 |
| A | 8 | 315.97 | 37,977.88 | 3.078 | 41.896 | 3.165 | 0.145 | 17.461 |
| B | 8 | 518.70 | 23,134.63 | 2.767 | 26.622 | 1.928 | 0.126 | 17.137 |
| C | 8 | 330.26 | 36,334.47 | 3.321 | 40.477 | 3.028 | 0.164 | 18.195 |
| A | 9 | 311.42 | 38,533.43 | 3.059 | 42.427 | 3.211 | 0.136 | 17.383 |
| B | 9 | 538.02 | 22,304.03 | 2.690 | 25.719 | 1.859 | 0.132 | 12.247 |
| C | 9 | 334.08 | 35,919.34 | 3.105 | 39.837 | 2.993 | 0.169 | 18.280 |
| A | 10 | 321.04 | 37,379.03 | 3.081 | 41.291 | 3.115 | 0.149 | 16.942 |
| B | 10 | 545.84 | 21,984.49 | 2.682 | 25.389 | 1.832 | 0.127 | 12.595 |
| C | 10 | 337.63 | 35,542.23 | 3.087 | 39.445 | 2.962 | 0.159 | 17.518 |
| A | 11 | 314.18 | 38,195.05 | 3.082 | 42.121 | 3.183 | 0.140 | 17.212 |
| B | 11 | 540.56 | 22,199.03 | 2.679 | 25.596 | 1.850 | 0.132 | 12.855 |
| C | 11 | 336.10 | 35,703.44 | 3.152 | 39.672 | 2.975 | 0.168 | 17.968 |
| A | 12 | 319.88 | 37,513.77 | 3.060 | 41.400 | 3.126 | 0.151 | 17.155 |
| B | 12 | 531.38 | 22,582.68 | 2.703 | 26.016 | 1.882 | 0.130 | 12.806 |
| C | 12 | 339.34 | 35,362.78 | 3.117 | 39.287 | 2.947 | 0.154 | 18.892 |
| A | 13 | 321.31 | 37,346.81 | 3.062 | 41.237 | 3.112 | 0.152 | 17.243 |
| B | 13 | 538.71 | 22,275.42 | 2.706 | 25.693 | 1.856 | 0.131 | 13.338 |
| C | 13 | 337.05 | 35,602.55 | 3.079 | 39.494 | 2.967 | 0.168 | 17.877 |
| A | 14 | 317.14 | 37,838.61 | 3.050 | 41.714 | 3.153 | 0.139 | 17.443 |
| B | 14 | 544.46 | 22,040.21 | 2.700 | 25.462 | 1.837 | 0.130 | 12.666 |
| C | 14 | 333.62 | 35,969.57 | 3.133 | 39.925 | 2.997 | 0.166 | 18.458 |
| A | 15 | 317.86 | 37,752.52 | 3.060 | 41.645 | 3.146 | 0.138 | 17.213 |
| B | 15 | 544.90 | 22,022.50 | 2.676 | 25.401 | 1.835 | 0.128 | 12.710 |
| C | 15 | 333.76 | 35,954.36 | 3.104 | 39.874 | 2.996 | 0.162 | 17.250 |
| A | 16 | 315.29 | 38,060.41 | 3.056 | 41.944 | 3.172 | 0.140 | 16.960 |
| B | 16 | 529.43 | 22,665.76 | 2.676 | 26.094 | 1.889 | 0.126 | 12.624 |
| C | 16 | 339.33 | 35,364.00 | 3.191 | 39.364 | 2.947 | 0.169 | 19.228 |
| A | 17 | 321.82 | 37,287.77 | 3.039 | 41.160 | 3.107 | 0.142 | 18.777 |
| B | 17 | 537.02 | 22,345.34 | 2.666 | 25.721 | 1.862 | 0.121 | 12.597 |
| C | 17 | 329.95 | 36,369.14 | 3.109 | 40.300 | 3.031 | 0.161 | 17.766 |
| A | 18 | 317.86 | 37,752.56 | 3.060 | 41.644 | 3.146 | 0.144 | 17.069 |
| B | 18 | 545.82 | 21,985.43 | 2.682 | 25.379 | 1.832 | 0.127 | 12.836 |
| C | 18 | 340.30 | 35,263.21 | 3.130 | 39.212 | 2.939 | 0.166 | 17.728 |
| A | 19 | 321.19 | 37,361.27 | 3.084 | 41.294 | 3.113 | 0.148 | 17.163 |
| B | 19 | 554.11 | 21,656.33 | 2.698 | 25.070 | 1.805 | 0.127 | 12.879 |
| C | 19 | 338.68 | 35,431.45 | 3.098 | 39.335 | 2.953 | 0.156 | 17.940 |
| A | 20 | 317.62 | 37,781.08 | 3.059 | 41.672 | 3.148 | 0.148 | 16.935 |
| B | 20 | 532.37 | 22,540.89 | 2.684 | 25.936 | 1.878 | 0.132 | 12.441 |
| C | 20 | 337.65 | 35,539.73 | 3.086 | 39.443 | 2.962 | 0.172 | 18.168 |
| A | 21 | 322.06 | 37,260.38 | 3.049 | 41.150 | 3.105 | 0.143 | 17.456 |
| B | 21 | 547.28 | 21,926.62 | 2.679 | 25.315 | 1.827 | 0.131 | 12.626 |
| C | 21 | 336.77 | 35,633.02 | 3.133 | 39.584 | 2.969 | 0.162 | 18.813 |
| A | 22 | 315.80 | 37,999.15 | 3.119 | 41.959 | 3.167 | 0.146 | 22.917 |
| B | 22 | 547.15 | 21,931.77 | 2.689 | 25.335 | 1.828 | 0.130 | 13.926 |
| C | 22 | 339.53 | 35,342.64 | 3.100 | 39.256 | 2.945 | 0.165 | 18.333 |
| A | 23 | 322.10 | 37,255.75 | 3.059 | 41.149 | 3.105 | 0.147 | 16.916 |
| B | 23 | 536.14 | 22,382.26 | 2.681 | 25.787 | 1.865 | 0.134 | 12.652 |
| C | 23 | 337.33 | 35,573.49 | 3.134 | 39.531 | 2.964 | 0.161 | 18.005 |
| A | 24 | 320.92 | 37,392.98 | 3.084 | 41.324 | 3.116 | 0.146 | 17.134 |
| B | 24 | 541.44 | 22,163.29 | 2.683 | 25.572 | 1.847 | 0.127 | 13.113 |
| C | 24 | 328.84 | 36,492.24 | 3.105 | 40.417 | 3.041 | 0.159 | 18.007 |
| A | 25 | 312.61 | 38,386.11 | 3.077 | 42.307 | 3.199 | 0.157 | 17.217 |
| B | 25 | 532.75 | 22,524.48 | 2.705 | 25.947 | 1.877 | 0.132 | 12.310 |
| C | 25 | 339.16 | 35,381.72 | 3.105 | 39.308 | 2.948 | 0.159 | 17.794 |
| A | 26 | 319.93 | 37,507.94 | 3.078 | 41.426 | 3.126 | 0.145 | 17.312 |
| B | 26 | 537.89 | 22,309.34 | 2.682 | 25.712 | 1.859 | 0.132 | 12.351 |
| C | 26 | 339.64 | 35,331.72 | 3.116 | 39.269 | 2.944 | 0.169 | 17.766 |
| A | 27 | 320.98 | 37,385.71 | 3.115 | 41.335 | 3.115 | 0.150 | 17.214 |
| B | 27 | 524.58 | 22,875.54 | 2.673 | 26.268 | 1.906 | 0.134 | 12.742 |
| C | 27 | 337.72 | 35,532.58 | 3.134 | 39.500 | 2.961 | 0.170 | 17.401 |
| A | 28 | 318.37 | 37,692.36 | 3.046 | 41.584 | 3.141 | 0.145 | 17.342 |
| B | 28 | 534.95 | 22,432.04 | 2.718 | 25.867 | 1.869 | 0.130 | 12.473 |
| C | 28 | 334.19 | 35,907.70 | 3.108 | 39.835 | 2.992 | 0.162 | 18.364 |
| A | 29 | 313.02 | 38,336.14 | 3.106 | 42.281 | 3.195 | 0.136 | 17.406 |
| B | 29 | 546.15 | 21,972.06 | 2.699 | 25.386 | 1.831 | 0.132 | 12.219 |
| C | 29 | 338.19 | 35,483.40 | 3.128 | 39.433 | 2.957 | 0.164 | 18.804 |
| A | 30 | 322.34 | 37,228.00 | 3.047 | 41.114 | 3.102 | 0.140 | 17.110 |
| B | 30 | 542.30 | 22,128.05 | 2.690 | 25.530 | 1.844 | 0.131 | 12.348 |
| C | 30 | 329.33 | 36,438.00 | 3.164 | 40.432 | 3.036 | 0.160 | 19.835 |